Long-Term Study of a Portable Field Robot in Urban Terrainhic/Georgia-HomePage/Publishing...Wiley...

Wiley author proof [ROB-07-0014_20214]

Wiley author proof [RO

B-07-0014_20214]

Long-Term Study of a PortableField Robot in Urban Terrain

Carl LundbergNational Defence College115 93 Stockholm, Swedene-mail: [email protected]

Henrik I. ChristensenGeorgia Institute of TechnologyAtlanta, Georgia 30332e-mail: [email protected]

Roger ReinholdLänna Hammarby 6945761 93 Norrtälje, Swedene-mail: [email protected]

Received 9 January 2007; accepted 23 July 2007

The armed forces have a considerable amount of experience in using robots for bombremoval and mine clearing. Emerging technology also enables the targeting of other ap-plications. To evaluate if real deployment of new technology is justified, tactical advan-tages gained have to be compared to drawbacks imposed. Evaluation calls for realistictests which in turn require methods dictating how to deploy the new features. The presentstudy has had two objectives: first, to gain a comprehensive view of a potential user ofman-portable robots; second, to embed a robot system with users for assessment ofpresent technology in real deployment. In this project we investigated an army companyspecialized in urban operations performing their tasks with the support of the iRobotPackbot Scout. The robot was integrated and deployed as an ordinary piece of equipmentwhich required modifying and retraining a number of standard behaviors. The reportedresults were acquired through a long-term test ranging over a period of six months. Thispaper focuses on the characteristics of the users and their current ways of operation; howthe robot was implemented and deployed. Additionally, this paper describes benefits anddrawbacks from the users’ perspective. A number of limitations in current robot tech-nology are also identified. The findings show that the military relies on precise and thor-oughly trained actions that can be executed with a minimum of ambiguity. To make useof robots, new behavioral schemes, which call for tactical optimization over several years,

FIELD REPORT

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Journal of Field Robotics 24(8/9), 1–XXXX (2007) © 2007 Wiley Periodicals, Inc.Published online in Wiley InterScience (www.interscience.wiley.com). • DOI: 10.1002/rob.20214




B-07-0014_20214] are needed. The most common application during the trials was reconnaissance insidebuildings with uncertain enemy presence when time was not critical. Deploying the robottook more time than completing the task by traditional means, but in return kept thesoldiers out of harm’s way and enabled them to decrease weapon deployment. The rangeof the radio link, limited video feedback, and the bulky operator control unit were thefeatures constraining the system’s overall performance the most. On the other hand, didproperties of the system, such as ruggedness, size, weight, terrain ability, and endurance,prove to match the application. The users were of the opinion that robots such as thePackbot Scout would be a valuable standard feature in urban intervention. © 2007 WileyPeriodicals, Inc.

1. INTRODUCTION

Mobile robots, also referred to as Unmanned GroundVehicles �UGVs�, have been used extensively for Ex-plosives Ordnance Disposal1 �EOD� and mine clear-ance operations for quite some time. Military, police,fire brigade, and rescue services have recently startedto evaluate robot technology as an aid to remove hu-mans from unnecessary harm, to perform more effi-ciently or at lower cost, and to enable missions un-suitable for humans. The question of when to acquirenew technology for real deployment is complex andinterdisciplinary. Possible advantages have to beweighed against costs for acquisition, integration,training, and maintenance, as well as mission effi-ciency and reliability. The work presented in this pa-per addresses the cost-benefit analysis from the per-spective of a selected end user—military in urbanintervention. Issues investigated have been such thatthe chosen users were competent to respond to. Thishas excluded the evaluation of, for example, costs ineconomical terms2 or valuing the humanitarian andpolitical benefits of avoiding casualties.

Examining the pros and cons experienced by us-ers, first of all, requires access to their arena. Thismight, due to reasons of safety and security, be a deli-cate matter when targeting high-risk work-groups.Second, user assessment demands a thorough knowl-edge of the users and their current work schemes asthis constitutes the baseline for comparison. Evaluat-ing new features calls for the development of meth-ods for deployment. Since current methods are highlyoptimized, users need to be adequately trained forthe new behaviors before comparison can take place.

The present study has had two objectives. Thefirst has been to gain a comprehensive view of a po-tential user of man-portable robots. This includes task

analysis, investigation of user characteristics, andidentification of limiting factors of today’s methods.The second objective has been to embed a robot sys-tem with the chosen user for assessment of presenttechnology in real deployment. This includes identi-fication of suitable robot missions, evaluation of us-ability, and comparison of pros and cons that robotdeployment involves. Emphasis has been placed onperforming the study under as realistic circumstancesas possible. Only technology that could be operatedby the end users in their ordinary environment wasregarded since the validity of users’ opinions con-cerning what could be tested in reality differs vastlyfrom speculations regarding features that might beincluded on future versions.3 Analogously, the resultsof the study were divided into two groups:

1. Results based on methods and technologythat were tested during the trials. Achievingsuch results demands both having hardwarethat can represent the investigated feature ina realistic way, and that the feature is used of-ten enough to be studied with validity.

2. Results concerning technology or methodswhich were not evaluated in depth. The rea-son for this could be either that the technol-ogy was not yet available, or that the featurein question was not deployed often enoughto be statistically verified with adequateconfidence.

This paper primarily presents the results whichfall into the first category. Initial findings withinthe second category have been previously re-ported �Lundberg, Reinhold & Christensen,2007b�.As this was the first step in the investigation of

1Removal, disarmament, and destruction of explosives.2Soldiers and officers in the field are not aware of costs for acqui-sition, training, logistics, or maintenance.

3The same applies to speculations about what is needed by othertypes of users than those who actually participated in the tests.

2 • Journal of Field Robotics—2007

Journal of Field Robotics DOI 10.1002/rob




B-07-0014_20214] robot deployment amongst the selected users, the ap-proach for research has, to a large extent, been per-formed through methods of anthropology such as ob-servation and interviews. Quantitative data mainlyderived from a questionnaire at the end of the trials.The three data collection methods triangulate4 theoutcome of the study which, to a large extent, is quali-tative rather than quantitative. Some of the conclu-sions drawn primarily verify what has been an ap-parent problem in the arena for a long time. In othercases we believe to have proved that available tech-nology is now able to match the users’ requirements.

The present work is intended to assist technicalresearchers and developers of field robots by provid-ing a holistic view of a user with high expectations ofwhat can be achieved with robots. Further, this workis directed towards those dealing with acquisitionand tactical development concerning robots for high-risk work forces. The work performed will also serveas the necessary basis for the design of coming in-depth experiments.5

This project has been a joint initiative between theRoyal Institute of Technology �KTH�, the NationalDefence College �FHS�, the Swedish Defence MaterielAdministration �FMV�, and the Royal Life Guards�LG� of the Swedish Armed Forces. An infantry com-pany, specialized in Military Operations in UrbanTerrain �MOUT�, constituted the user group for thestudy. The company provided full access for researchwhile using an iRobot Packbot Scout on their trainingmaneuvers during a period of six months.

The remainder of the paper is organized in thefollowing way: Section 2 contains a discussion of re-lated work. Section 3 presents a description of the ro-bot, the user, the test facilities, and how implemen-tation and research was performed. Section 4 containsthe results, divided into three categories: Subsection4.1 presents characteristic of the users while Subsec-tions 4.2 and 4.3 describe the tactical and the technicalfindings of robot use, respectively. Section 5 gives adiscussion of the project, the research methods, aswell as the main results.

2. RELATED WORK

Various studies have previously investigated high-risk workers deploying field robots. The most com-mon application, bomb destruction or removal, hasbeen successively refined since the first attempts inNorthern Ireland in the beginning of the 1970’s �Bir-chall, 1997�. Today this is a well-established robotniche with several mature systems available as dem-onstrated at the European Land-Robot Trial 2006�Schneider, 2007�. Other areas of robot deploymentshared by the police and military and given substan-tial investments are security, surveillance, reconnais-sance, and tactical support �Carroll, Nguyen, Everett& Frederick, 2005; Jones, Rock, Burns & Morris, 2002;Kumagai, 2002; Barnes, Everett & Rudakevych, 2005;Ebert & Stratton, 2005; SPAWAR, 2007�, althoughmuch of the research is not published in detail �Ash-ley, 2006�.

The task of chemical, biological, radiological, andnuclear �CBRN� contamination control is about to beaddressed as sensor payloads are maturing for de-ployment on EOD-robots in daily use, such as thePackbot, Talon, Wolverine, & URBOT �Smith-Detection, 2007; Foster-Miller, 2007; Gardner, Treado,Jochem & Gilbert, 2006; SPAWAR, 2007�. Rescue ro-botics, and especially Urban Search and Rescue�USAR�, is one of the field robot areas currently re-ceiving the most attention in research studies. Coun-termeasures against, and preparedness for terroristattacks and earthquakes have invigorated efforts topush robot technology into use �Murphy, 2004; Hisa-nori, 2002; Matsuno & Tadokoro, 2004; Scholtz,Young, Drury & Yanco, 2004; Yanco & Drury, 2004�.

Human-robot interaction outside the scope ofhigh-risk field workers has been evaluated in long-term tests as well. An early example is the integrationof the SURBOT �White, Harvey & Farnstrom, 1987�for mobile surveillance in a nuclear power plant.More recent examples consist of testing of the robotseal Paro amongst elderly �Wada, Shibata, Saito,Sakamoto & Tanie, 2005�, of the fetch-and-carry robotCERO by a partially impaired person �Hüttenrauch &Eklundh, 2002� and a number of long-term tests oftour guide robots such as the RoboX9 at Expo02�Tomatis, Terrien, Piguet, Burnier, Bouabdallah, Arras& Siegwart, 2003�. By now space applications havealso been trialled substantially through NASA’s de-ployment of rovers on Mars �Leger, Trebi-Ollennu,Wright, Maxwell, Bonitz, Biesiadecki, Hartman, Coo-per, Baumgartner & Maimone, 2005�.

4Triangulation of data see �Silverman, 2006�.5One such experiment has already been performed to comparemanual mapping with teleoperation mapping �Lundberg & Chris-tensen, 2006�.

Lundberg et al.: Long-Term Study of a Portable Field Robot in Urban Terrain • 3





B-07-0014_20214]

The above-mentioned examples have in commonthe implementation in a realistic setting over a periodof time. Despite the similar aim of these projects, ap-proaches may vary significantly depending on objec-tive, resources, and type of application. For example,evaluation amongst elderly will differ vastly fromevaluation with search and rescue personnel. Like-wise, technical requisites to enable testing varygreatly between, for example, deployments in a mu-seum and on Mars. Hence, there is no obvious anduniform way to evaluate the deployment; establishedresearch methods may not meet the requirements toevaluate the telepresence, dynamics, or autonomy ofmobile robots �Scholtz et al., 2004�. Traditionalmethods—developed for other fields such as human-computer interaction, home electronics, social an-thropology, or psychology—need to be tested, per-haps also modified and merged, in order to facilitateevaluation of robotics �Thomas & Macredie, 2002�.The application targeted in this project is distin-guished from previous work on risky applications byits highly dynamic and responsive target.6 From thisperspective it has more in common with more casualapplications such as service robotics. The project hasfielded a commercially available robot in an arenathat is, for reasons of secrecy, more frequently re-searched than published. Research taking a compre-hensive view of robot deployment within MOUT hasnot been previously reported.

3. METHODOLOGY AND SETTING

3.1. Outline of Study

A pre-study initiated the project with the purpose toinvestigate the user and to explore in which ways

they could benefit from using robots �Lundberg,Christensen & Hedstrom, 2005�. The pre-studyimplemented and performed exploratory testing ofthe robot during five military training maneuversJanuary to March 2005. The robot was deployed byconscript soldiers and tested within squad level dur-ing four maneuvers. On the final maneuver, one ofthe researchers, who is also an officer, operated therobot in order to deploy on company level. Pre-study data was gathered through military manuals,field observation, formal and informal interviews,and participatory observation �Figure 1�.

The pre-study led to a decision to implementlong-term testing on a larger scale during the follow-ing year. A selection of standard procedures was re-designed to include the robot. Once mastered by theoperators, the new procedures were demonstrated tothe rest of the company. It was thereafter up to thecommanders of the company to use the robot as theyconsidered appropriate in their training maneuvers.

3.2. User Group

The Swedish Armed Forces are based on conscrip-tion of 18 to 24 year old males. The Army uses con-script soldiers in positions up to second platooncommander. Higher positions are held by profes-sional officers. Women can volunteer to do militaryservice, and all positions within the Swedish ArmedForces are open to both genders. Basic military train-ing service ranges from eight to fifteen months. Stan-dard units such as the company in our study aretrained in one-year cycles.

The appointed MOUT-company consisted of 200soldiers, 7 professional officers, 16 Armoured Per-sonal Carriers7 �APCs�, and 10 logistic vehicles.

6Enemy soldiers or civilians instead of artifacts such as bombs orhazardous chemicals.

7Tracked vehicles, armed with 20 mm cannons, capacity to carry11 people �Hägglunds Vehicle AB, PBV302�.

Figure 1. The various data collection methods deployed during the three phases of the test.






B-07-0014_20214] About three percent of the company members werefemale. Ten additional officers served as instructors.Soldier training initially targets individual behav-iors; complexity of training is gradually increased onsquad, platoon, company, and, finally, battalionlevel. Lower-level training receives more time ashigher-level training requires larger logistical, per-sonnel, and organizational resources. MOUT-unitsfurther require access to urban surroundings suit-able for training. This includes moving entire unitsnationwide, or even to neighboring countries, in or-der to access appropriate exercise areas.

Performing tests on a specific organizationallevel calls for the framework of at least one levelhigher, i.e., testing squad behavior requires theframework of a platoon, and platoon assessment inturn calls for a realistic company setting. For ex-ample, the majority of participants provided the con-text for the test rather than interacting with theproduct.

During the six month test period, the companyacted as a cohesive unit on five training maneuverswhich included 200 to 6000 soldiers and lasted forthree to six days.8 The maneuvers were set up withan initial phase of political conflict, mobilization,and handover of tactical responsibility to the com-pany. The robot had no role during these initialphases which, generally, lasted for one-half to oneday. The robot, however, entered the scene duringcombat for a total of 16 days.

The company under study is a highly special-ized unit which directs about 90% of its training to-wards urban intervention. Ordinary troops have anopposite distribution with open terrain as their pri-mary area of operation. Many current conflicts, how-ever, take place in urban environments, and a corre-sponding redirection of armed force training isongoing.

3.3. Test Facilities

All tests were carried out in facilities regularly usedfor military training. Deserted and partly demol-ished industrial and residential buildings offered anenvironment similar to those expected during com-bat operations �Figure 2�. During the tests no adap-

tations or adjustments were done to the environ-ment. The test period spanned from autumn tospring and temperatures from −15°C to +15°C.

3.4. Robot System

The iRobot Packbot Scout9 �Figure 3� used in thestudy was equipped with a number of accessories.The same Direct Fire Weapon Effects Simulator �DF-WES� that is used for combat training was mountedon the robot �Saab BT46�. The system includes a la-ser mounted on the firearms, and a sensor suit wornby the soldiers. Adding the fire simulation systemenabled the soldiers to engage the robot with theirfirearms during training.

Two more payloads were developed; a flashlight

8The maneuvers took place in Fagersta 1–4 Nov. 2005, Enköping8–9 Nov. 2005 �Demo 05�, Marma 5–9 Dec. 2005, Fagersta 6–9 Feb.2006, Jordberga 5–7 March 2006, Norrköping/Stockholm 20–26March 2006.

9See �Lundberg et al., 2005; iRobot, 2007� for a description of thePackbot Scout.

Figure 2. One of the test environments, a deserted steelfactory complex in Fagersta. To seize such a building is atypical task for a company.

Figure 3. The Packbot Scout equipped with the Claymoremine �top�, and the Direct Fire Weapon Effects Simulator.In front, the remote control for payload operation.






B-07-0014_20214] for illumination in dark premises and a blank Clay-more mine, both triggered by remote control. A sirenon the robot simulated detonation of the Claymoremine. The choice of weapon was not a result of de-tailed consideration. A Claymore mine, which is astandard weapon for infantry troops, would mostcertainly destroy the robot if detonated. The minemight also be a much too powerful weapon for in-door use or while in the vicinity of our own person-nel. It does not, however, have to be aimed very ac-curately. Nonlethal weapons such as flash-banggrenades might be a more suitable option for initialimplementation.10

Along the course of the test, the system also in-corporated extra batteries, chargers, basic spareparts, a rope for lowering the robot into premises, acarrying system, and protective cases for all compo-nents. A telescopic rod, which could be attached ver-tically to the robot to detonate trip-wired explosives,was also included.

3.5. Robot Implementation

During the pre-study, the robot had not been in com-pany hands except during the individual trial ap-pointments. For the main study test, except duringmodifications and repairs, the robot was kept by thecompany. All transports, maintenance, and chargingwas carried out with the test groups’ ordinary re-sources in order to expose the system to realisticstress. The users were instructed to treat the robot asa standard piece of equipment and they were toldthat damage and wear was considered an expectedside effect of the study. The robot’s robustness wasstated to be comparable to the users’ radios or opti-cal equipment.

In agreement with the Commander of the RoyalLife Guards, one officer and two soldiers weretrained to operate the robot during the main trials.Unfortunately, one of the soldiers was released fromduty due to medical reasons after two months. Theoperator task was performed according to the regu-lar rules and demands of the unit.

No military tactical concepts of robot deploy-ment existed when the robot was first introduced tothe company, nor did the users have any well-

defined opinions about how the system could beimplemented. This was the first acquaintance withrobotics for most of them. The delegated operatorshad no previous experience with robotics, but wereaccustomed computer users and had some experi-ence with RC-crafts.

Three levels of operator training were defined: 1.Basic Level, 2. Map and Search Level, and 3. TacticalLevel. Training for the Basic Level followed thescheme developed during the pre-study. This in-cluded briefing about the robot, basic driving, andfamiliarization with the appearance through videofeedback. After the basic training, which took oneday, the operators were able to perform simple mis-sions and continue to practice on their own. Sincethe higher level behaviors were not yet defined, thehigher level training was, to a large extent per-formed concurrent with exploratory testing. TheMap and Search Level incorporated the ability tosketch explored premises and to search for personsor Improvised Explosive Devices �IEDs�.

After having acquired personal skills in robotcontrol, the operators were trained to execute mis-sions in conjunction with others, i.e., the TacticalLevel. Initially, this was done in pairs, where the op-erator and an assisting soldier were trained to act asa team �Figure 4�. After approximately seven days ofpractice,11 the pair was integrated into group, squad,and platoon level, and performed their tasks syn-chronized with other mission activities �Figure 5�.While the two first training levels only included theoperator and the assistant, the third level also re-quired adaptation of the group in which the robotoperated.

Once the operators had acquired the necessaryskills, demonstrations were done for one platoon ata time and included a briefing about the system,safety issues, and a demonstration of an explorationmission. It was emphasized that the robot use was atest and that some aspects could not be expected toreach full effectiveness until after some time of tac-tical development and training. After the demo thecompany was free to use the robot system as theypleased. During the training maneuvers, either theDFWES-system12 or the training officers could deter-mine the robot had been neutralized; if so, the robotwas returned to the transportation vehicle from10The use of armed robots is not a new occurrence. EOD-robots

have been equipped with disruptor guns for more than three de-cades �Birchall, 1997�, and attempts to use robots for explosivedeliverance date as far back as World War II �Chamberlain, Doyle& Jentz, 1993�.

11As a comparison, the basic operator course for EOD-technicianslasts five days �Lundberg, Reinhold, & Christensen, 2007a�.12See Section 3.4.






B-07-0014_20214]

where the operator could retrieve it as a new robot.There was no set limit for how many times thiscould be done, i.e., the company had a fictitiouslyunlimited resource of robots. The company had ac-cess to one robot system on all maneuvers except forthe 2006 final maneuver when two systems were de-ployed in parallel.

Prior to testing with conscripts, military safetyregulations required the system to be reviewed bythe Armed Forces Safety Board, which identified therisk of the robot falling onto a person situated belowas the largest hazard. All participating personnelwere informed about robot safety prior to everytraining maneuver.

3.6. Data Collection and Analysis

Military training manuals and videos proved valu-able tools to get familiarized with the users’ circum-stances, their terminology, and their work schemes�Figure 1�. In addition, participating in one

national13 and one international workshop14 on ur-ban warfare presented the opportunity to investigateareas of tactical development.

Observation, with video and photography docu-mentation, was an essential research method for theproject. The users are accustomed to the presence ofinstructors and visitors during exercises. Armlets areused to distinguish observers from participants,which allowed for full access observation during themaneuvers. For safety reasons, hearing protectionand protective goggles had to be worn by militarypersonnel as well as civilians while indoors duringthe firing of blank ammunition. Radio communica-tion between the platoon leaders and the companycommander added an alternative method of obser-vation. This is an established way for training offic-ers to stay informed about the units’ overallprogress. Observing the process from the enemy side

13The annual MOUT-development workshop of the Army CombatSchool, Borensberg, 10 May 2005.14Trilateral workshop on Urban Warfare, The Ministries of De-fence of Germany �FhG-INT�, The Netherlands Defence ResearchInstitutes �TNO�, Swedish Defence Research Agency �FOI�, andSwedish Defence Materiel Administration �FMV�, 18–19 May2005, Stockholm.

Figure 4. From left to right: the Packbot Scout, the robotassistant, and the robot operator. The robot is about tomove up the staircase to the left; meanwhile, the assistantis ready to act in the most hazardous direction, in this caseconsidered to be the staircase. The operator controls therobot from a previously secured area. Observing the robotin line of sight provides better situational awareness thanhaving to rely on the user interface. The assistant, there-fore, supports the operator with voice commands as longas he has the robot in sight.

Figure 5. From the left: the platoon leader, the robot op-erator, the group leader, and thereafter the private sol-diers. The platoon leader is using the robot-system to per-form exploration around the corner. The group leader isobserving the neighborhood and the soldiers are ready toact in hazardous directions.






B-07-0014_20214] provided a beneficial alternative point of view. Re-searchers also took the place of robot operator forparticipatory observation purposes.15

Three types of interviews were conducted dur-ing the project. First, two officers were interviewedbefore the long-term test regarding what applica-tions they thought might be feasible for robot imple-mentation. The results were verified by the respon-dents after transcription. Second, participants wereinterviewed about their established procedures andtheir experience working with the robot. These inter-views were done spontaneously at appropriate timesin the field, and were held in an informal manner. Insome cases these interviews were documented withvideo or notes, but mostly the data was writtendown at a later time. Finally, after the long-termstudy, four soldiers and six officers were chosen forin-depth interviews regarding their experiencesfrom using the robot. An anthropologist who hadnot participated in the field studies was recruited toperform the final interviews. These interviews,which lasted anywhere from 30 min to one houreach, were recorded and transcribed.16 The topics ex-plored in the final interviews were aligned with thetopics of the end-of-trial questionnaire. The semi-structured interviews provided the opportunity toextend and modify the questions for in-depth inves-tigations.

The 41 most experienced participants �36 sol-diers and 5 officers� were selected for the question-naire study. The inquiry contained 14 main ques-tions which were either statements to be rated on afive point scale, or open-ended questions. The re-spondents were given the option to add alternativesthey believed to be missing. Participants took from20 to 60 min to complete the questionnaire. Thepost-evaluation �interviews and questionnaire� wasdesigned to document which robot missions hadbeen performed, explore the opinions about the ro-bots efficiency/functionality and its significancecompared to other equipment, probe ideas for futuredevelopment and ethical considerations, and inves-tigate whether acquisition was recommended.

4. RESULTS

4.1. User Characteristics

Military operations in urban terrain might be one ofthe most challenging team tasks performed. Com-plexity, lack of information, personal risk, fatigue,and time pressure in possible combination with limi-tations in experience make the work environmenthighly demanding. Forces deployed in peacekeepingand peace enforcement missions might also be pres-sured by cultural differences, media, and ethicaldoubts. Compared to missions in field settings, mis-sions in urban terrain often tend to be more frag-mented without clear borders between civilians, en-emies, and allies �Krulak, 1999�. The numerouspossible hideouts and the difficulty to overview ur-ban areas call for a large number of personnel tocontrol an area even against small enemy forces. Forexample, to seize and secure a building such as theone displayed in Figure 2 would be a typical task fora MOUT-company.

A company can be described as the smallest self-sufficient army unit in the sense that it has medicalresources and supplies to last a couple of days, canhandle vehicle towing and field repairs, and is giventactical responsibility for a defined area. Commandis carried out by two captains with six to ten years ofexperience. Only one is the formal commander, butin urban operations they often work in unison, oneleading the soldiers moving on foot while the othermanages the Armoured Personnel Carriers �APCs�.A MOUT-Company consists of three fighting pla-toons and a Staff/Supply platoon. Each platoon iscommanded by a lieutenant with three to six yearsof experience who is assisted by a conscript ser-geant. The platoon leaders and the assisting ser-geants carry radios for communication with thecompany commanders and the APCs. Each MOUT-platoon consists of four squads, each with an APC.Each soldier has an area of expertise such as squadleader, machine gunner, sharp shooter, demolitionexpert, combat medic, APC-commander, APC-gunman, or APC-driver. When moving on foot, thesoldiers perform in pairs �buddy-system� in order toprovide mutual protection, assist during obstaclepassing, and so on. Tasks that have to be solved inexposed settings are executed according to strict rou-tines. For example, soldiers having a weapon mal-function calls out “malfunction” and drops low inorder to take cover as well as to visually inform the

15Just as during the pre-study, one of the researchers acted asoperator during one maneuver when the trained operators wereunavailable.16Each interview took eight hours on average to transcribe andabout as long to analyze.






B-07-0014_20214] others about the inability to solve the current task.Meanwhile, the soldier’s “buddy” positions behindhim or her as a lookout in order to offer protectionuntil the problem has been resolved.

For transport in urban terrain the troops havethe option to use the APCs. In case of hostile en-counters the soldiers leave the vehicles and continuetheir movement on foot through buildings to besheltered and hidden. From then on all the necessaryequipment has to be carried, which requires all gearto be compact, light, and rugged. Only the most nec-essary equipment and limited amounts of suppliescan be brought along. Soldiers are, however, alreadyfully loaded and forced to leave important gear, sup-plies, and water behind.

The combat vehicles are often left behind be-cause they have limited mobility and constitute arelatively obvious target which, despite their armor,cannot withstand the firepower of modern portableanti-tank weapons and mines. Normally a driver, amachine-gunner, and an APC-commander remain inthe vehicle in order to perform medical evacuation,fire support, outflanks, etc. While moving on footthe soldiers have to be able to traverse obstructions,climb through windows and onto roofs.

Most indoor operations are carried out in duskor darkness with either helmet-mounted night vi-sion goggles or powerful flashlight illumination.Both have their drawbacks. The night vision gogglescognitively impair the wearer by limiting the field ofview and eye movement, removing stereo vision,and obstructing rifle deployment �Sandberg, 2003�.A flashlight, on the other hand, reveals the positionof the user. One possible solution, which takes thepros and cons of both into account, is to use thenight vision aids for stealth reconnaissance, and whitelight for armed action during which the ability ofperception and speed is prioritized.

Communication and distribution of informationis limited during urban operations. First, the circum-stances for voice communication are often poor dueto physical factors and the necessity to operate insilence. Second, there is an aggravating factor due tothe high stress level on the personnel. Third, the or-ganization is strictly hierarchical so all informationtransfer and most of the decisions are handledthrough the chain of command. In addition, some ofthe leaders have to communicate on two differentnetworks, one for subordinates and the other for su-

periors. However, units or individuals cut off fromcommunication will continue to act according to thecommanders’ outline plan.

To improve the information distribution, testsare currently done to equip every single soldier witha radio for inter-squad voice communication. The ra-dios have proved to eliminate many of the commu-nication problems related to the deliverance of ver-bal messages, but much of the information is spatialand difficult to present through speech. The use ofsketches is common when communicating face toface, and basic sign language is used for line of sightcommunication. Trials are also made to equip sol-diers with wearable computers for GPS-positioning,digital maps, aerial photographs or satellite images,digital cameras, and radio data networks for infor-mation sharing �Hoving, 2003�.

Despite communication difficulties the units areexpected to perform swiftly and in synchronization.Predefined tactics help the soldiers anticipate howtheir own forces will act in situations when planningor communication is lacking. All basic military be-haviors have been thoroughly defined and trained inorder to minimize reaction times, optimize effi-ciency, and minimize the risk of fratricide. For ex-ample, on squad-level it is defined in detail whatequipment each soldier carries, which task each sol-dier performs, how the squad moves, who opensdoors, who is the first to enter, how communicationis carried out, and so on. Established routines are notstatic but are continuously evaluated and refined.The evolution is, however, an incremental processperformed over many years. Examples of predefinedkey behaviors relevant when considering UGVs are:

• Reconnaissance–Gathering of informationthrough remote observation or by enteringthe specific area. Enemy encounter isavoided.

• Combat reconnaissance–Same purpose as forreconnaissance but with the aim to pursue intoattack in case of enemy encounter.

• Attack–Offensive action through maneuver-ing and firing of weapons in order to hold,neutralize, or force the enemy to surrender.

• Seize–Taking control of an area with or with-out enemy encounter. This includes ensuringthat the area is free from enemies and estab-lishing a defense.

• Search–Defensive way to perform seize. Sol-






B-07-0014_20214] diers go through a building room by roomwithout using any firepower. Search can beperformed silently.

• Cleanse–Offensive way to perform seize. Sol-diers go through a building room by roomwith preventive use of hand grenades andrifle fire, i.e., throwing grenades and shootinginto the next room before entering or even ob-serving it. Search often precedes cleanse untilenemies are encountered.

• Break-in–More or less aggressive action to en-ter a building. Obvious entrances such asdoors are avoided due to the risk of IEDs orambush. Ladders or vehicles are regularlyused to enter above-ground floor �Figure 6�.Armored vehicles can be used to breachthrough walls and provide massive fire sup-port to soldiers on foot. Break-ins are initiatedfrom a safe spot and are considered completedonce the troops are in control of a safe roominside the building. Performing a break-in is atask for an entire platoon and is considered tobe a very risky operation that has to be per-formed with rapid intensity. Break-ins areregularly preceded by diversions.

In addition to predefinition and training of fixedbehaviors, the MOUT-doctrine strongly emphasizesthat only one task is conducted at a time. Individualsoldiers or units are given only one objective at atime to avoid any ambiguity or mishaps due to men-tal overload. The high risk, uncertainty, and time

pressure has further brought the MOUT-troops to bevery conscious about reliability, a phenomena alsoobserved among police SWAT-teams �Jones et al.,2002�. Being able to execute a task according to planis favored over attempting something more ad-vanced that might not work. Failure to carry out anoutlined plan is considered a major tactical setbacksince one of the foundations of the military doctrineis to be proactive rather than reactive to the enemy’sactions. In fact, reliability along with the desire tomove and respond to new situations swiftly, is oftenconsidered more important than reducing risk �atleast during training maneuvers�. Often a task is sotime-constrained that if it cannot be solved withincertain limits, it does not need to be performed at all.

For international missions the enemy is expectedto be technically less well-equipped and trained butpossibly more experienced. Recent conflicts show anincreased use of asymmetric measures or means vio-lating international law, e.g., deploying snipers andsuicide bombers, resorting to terror actions, and tar-geting civilians �Krulak, 1999�. It is a well-established tactical fact that a defending force is bet-ter motivated and more willing to take risk, and alsoholds the advantage to fortify which renders apower balance of approximately one to four com-pared to the attacking force.

Likely tasks in international missions differ fromthose performed during soldier training. Despite aserious attitude and sometimes even frightful real-ism during exercise, it is hard to reconstruct the trueimpact of casualties. The aim during training is toget the most possible out of available resources andtime. Training is, therefore, directed towards com-plex tasks, such as offensive, high-paced, and full-scale battles performed in order to engage and,thereby, train, as many soldiers as possible. Routineduties such as surveillance or low-intensity conflictsreceive less attention. During the study the exam-ined MOUT-company normally trained two ratheroffensive seize missions per day, each rendering 5–30of their own wounded or killed �Figure 7�. This isnot a likely level for international missions. Troopsparticipating in international missions are reluctantto risk their own or civilian losses; a phenomenonthat has also been pointed out in other countries�Sion, 2006�. The need for training with less aggres-sive behaviors and nonlethal weapons is recognized.

The observed training maneuvers did not dealwith the dilemma of IEDs or anti-personnel mines toany large extent; although such threats might be a

Figure 6. Break-in performed through a window to avoidambush or IEDs. One squad is entering �right�, one isready to go �center�, and another is being briefed �left� bythe platoon leader �pointing�. The first goal after enteringis to establish a safe room.






B-07-0014_20214]

significant problem in international missions. Highambitions to obey international law during soldiertraining might be the reason for this.17

4.1.1. User Characteristics in Summary

MOUT are complex team tasks that require high lev-els of coordination. Time is often a critical issue andmeans of communication are sparse. The methods ofthe troops are highly defined and thoroughly prac-

ticed in order to enable synchronized actions evenwhen coordination is absent. The troops regularlyleave the APCs and continue their movementthrough buildings on foot. Alternative entrancessuch as windows are preferred to avoid ambush orIEDs. Weight is a critical issue as the soldiers arealready heavily loaded. The units are expected toperform their tasks no matter the type of premise,level of destruction, weather conditions, or time ofday. Massive weapon deployment and rapid actionis used to seize premises that are known to, or sus-pected to be, held by enemies. This increases the riskfor civilian casualties. Nonlethal weapons and meth-ods are desired although the training maneuvers areperformed more aggressively and at higher pacethan what is expected in international missions. Tac-tics are typically refined through small iterationsover long periods of time. The main user character-istics are also summarized in Table I.

4.2. Tactical Findings

4.2.1. Organization and Command

After the trials 83% of the users were of the opinionthat every MOUT-platoon should be equipped withone Packbot18 �i.e., three systems per company�.Having the robot systems as a standard part of theplatoon, instead of as a resource allowed access tofrom time to time, largely influences execution effi-ciency. It was considered that a MOUT-robot has tobe accessible within a few minutes; the tactical win-dow of opportunity does not allow for longer de-lays. The information gained with the robot is of

17Learning how to neutralize IEDs would also mean learning howto rig them.

18How many systems do you believe to be appropriate for a MOUT-company? � One per company? � One per platoon? � One per squad?

Figure 7. A medical log written on a wall in the safe room.This was the outcome of half a day of training maneuvers,the ones in the left column are the enemies, the ones in theright are their own. The log keeps track of the priority forevacuation of the wounded. Abbreviations: P1–very ur-gent, P2–urgent, P3–not urgent, P4–dead or mortallywounded, av/avtr/avtransporterad–evacuated.

Table I. User characteristics in summary.

Areas of concern Main features

Timeframe Often a very critical issueCommunication and synchronization Challenging conditions; behaviors are therefore, predefined and

well practicedOperational conditions Day and night, all weather and climateEnvironment Mainly urban, likely in destructionProperties of gear Rugged, man-portable, light weightOperational characteristics Today’s tactics are characterized as either rather defensive �prior to

interception� or highly aggressive �post-breakout of violence�; nonlethalalternatives are desired






B-07-0014_20214] most value to the soldiers about to enter the ex-plored area.19 As soon as any of the team membersenter the premises, they will gain a better controlthrough their own senses and means �weapons� thanwith the robot. Any information about enemy pres-ence is typically only valid for a short period of time,as revealed enemies are bound to change their posi-tion. The robot operator needs to be at the site ofexploration to grasp the overall setting in order tophysically handle the robot and be able to pass onthe information quickly.

Operating and handling the robot systemproved to be a two-person task due to both physicaland mental workload issues. To a large extent, themental workload imposed isolates the operator fromthe surroundings; this can, just as for other mentallydemanding tasks, be handled through the buddy-system. While the role of an assistant can be per-formed by any soldier given some additional train-ing, the operator requires at least one week ofindividual training.20 The operator needs to have thetactical and verbal skills of a squad leader in order tounderstand the tactical situation and to communi-cate gained information to others.

The number of persons required to deploy a ro-bot is often considered in the evaluation of systemefficiency �Murphy, 2004�. As mentioned, the mili-tary is accustomed to tasks that have to be per-formed in support of others, so the ratio of two per-sons per robot does not seem to be of criticalconcern. On the contrary, the MOUT-doctrine statesthat only one task should be solved at a time; ac-cordingly, it is not applicable to have one operatordeploying several systems simultaneously as is oftensuggested in research concerning autonomy �Jones& Lenser, 2006�.

The robot was one of the heavier pieces ofequipment handled; though occasionally, one soldierdrags or carries another wounded soldier whoweighs at least four times as much. The robot’sweight and size reduces the ability to perform dy-namic moves such as running and jumping. The ex-tra weight and size also cause problems while pass-ing narrows and crawling in cover. Furthermore,falling over with the extra weight increases the riskof injury. The robot crew needs to be physically

stronger, more motivated, and more alert than theaverage soldier while moving on foot. Although therobot and the Operator Control Unit �OCU� could becarried by one person, a pair moves with speed andendurance corresponding better to the rest of thetroops. The size and weight of the Packbot ap-proached the acceptable limit of MOUT; any weightreduction would be beneficial.

Similar to other company-shared resources, thecommander set a standard to submit the robot-system to the platoon momentarily having the high-est demand. Platoon leaders in charge of the robotsystem most commonly assigned it to one of thesquad leaders, who in turn acquired the information;alternatively, the leaders had the soldiers about toenter the explored area cooperate with the robot op-erator directly. After fulfilling a platoon’s needs, thesystem was released and relocated to the company’spost for medical evacuation to await the next mis-sion. Having the robot, and the operators, assignedto varying units also involved having them trans-ported by different APCs. This proved to cause lo-gistical problems since the vehicles were alreadyfully loaded with gear and supplies. Being able totransport and preferably launch the robot from theoutside of the APCs would be valuable.

4.2.2. Specification of Tactical Behaviors

A MOUT-scout robot needs to be closely integratedwith the deploying unit. Leaders on all levels, aswell as the individual soldiers, have to adapt andtrain their tactics to accommodate robot deploy-ment. Experience is required in order to accuratelydecide to which situations the UGV should be de-ployed; how the mission should be conducted; howlong a robot mission will likely last; what terrain therobot can handle; and what information the systemis capable of rendering. The trials meant to specifythe robot operation in as much detail as otherMOUT-behaviors. This included defining the opera-tor’s tasks as well as working out how squad-,platoon-, and company commanders should reasonregarding when and how to deploy the robot. Thelevel of detail is indicated by the following ex-amples:

• Operator level–Robot exploration should bedone along walls that provide a reference fornavigation. Deviations from walls should bedone in a perpendicular pattern; the compass

19Compared to, for example, UAVs, which in many cases are op-erated by others rather than those taking part of the informationgained by the system.20Covering the 1-Basic Level and 2-Map and Search Level describedin Section 3.5.






B-07-0014_20214] can be used for support although the risk oflocal deviation due to metal has to be consid-ered. The point at which the wall is tempo-rarily abandoned should be memorized to fa-cilitate return. This methodology is especiallyvital when exploring larger areas.

• Squad level–Since time will only allow for onereport, the exploration should be completedbefore passing on information; however, incase immediate progress is necessary, be pre-pared to interrupt the exploration to reportcurrent findings.

• Platoon level–The robot can be deployed fortasks that are normally performed withbackup from the rest of the platoon. This en-ables exploration beyond the point when allsoldiers are tied up guarding already seizedareas.

• Company level–The robot can be used forhigh-risk reconnaissance or for situations inwhich cleanse is not feasible due, for example,to the possible presence of civilians.

4.2.3. Executed Scenarios

During the training maneuvers the robot was de-ployed on two–ten occasions per day. Figure 8 dis-plays the distribution of the different types of mis-

sions according to the post questionnaire.21 Themissions reported most frequently were: reconnais-sance inside of buildings 37%, investigation ofbreak-in points 24%, and mapping 21%. The flash-light and Claymore payloads were deployed 8% and7%, respectively. When considering this distribution,it has to be taken into account that the payloadswere not available until the two last maneuvers.

4.2.4. Methods for Deployment

Robot reconnaissance was typically initiated from asafe spot where the squad could make a short stop incover. From the safe spot the squad or platoon leaderordered the operator to perform the search with spe-cifics on what area to explore, available time, andwhom to report to. It was beneficial to have the op-erator attend the company and platoon briefings togain general insight into the tactics. Attending thebriefings also enabled the operator to suggest how todeploy the UGV.

Robot exploration was done short steps at a timeto avoid losing radio contact and to ensure the ac-quired information was up to date. The tentative ex-ploration also enabled the commanders to keep upthe momentum; additionally, it reduced the need forsoldiers to memorize large amounts of information.The successive method was also motivated by thecommanders’ desires to perform immediate actiontowards enemies encountered, thereby minimizingthe forewarning effects of the robot. Consequently,exploration was done one or two rooms ahead andideally rendered a suitable safe spot for the soldiersto advance to.

Handing over spatial information is an impor-tant part of using the UGV for reconnaissance andmapping. The operator displayed sketches of all butvery simple premises on a small whiteboard, whichwas attached with Velcro on the lid of the OCU lap-top. The sketching was best done incrementally, 1–3walls or doors at a time. Attempts to keep too muchinformation in mind easily led to information loss.In the interest of time and clarity, only very basicinformation, such as walls, doors, and windows,

21Describe the different scenarios you have experienced during the UGV-trials…Select the type of scenario out of the following…: A. Reconnais-sance of Break-in point, B. Reconnaissance in rooms or corridors, C.Illumination with flashlight, D. Mapping, E. Deployment of Claymoremine, F. Surprise or mislead. �The response-fields were: Scenariotype �A–F�, number of times, which exercise area, course of event,pros and cons.�

Figure 8. The distribution �%� of the different scenariosreported in the questionnaire �see footnote 21�.






B-07-0014_20214] were depicted. When showing the sketch, the opera-tor added a few keywords describing the characterof the environment and any people observed.Sketching also helped the operator better grasp thelayout.

Nonoperators taking part of the image from therobot had difficulties assimilating the context of theexplored region. One likely reason is that passiveobservers do not have the operator’s motor motionof the joystick commands. Constant attention had tobe given to the robot’s camera view in order to un-derstand the spatial layout; this became clear duringtrials with multiple OCUs within the squad. If theoperator explained the robot’s current position withthe sketch, however, the camera view could be usedto point out specific observations. Unfortunately, theinterface design did not allow for snapshot imagesto be taken and stored for later viewing.

In the event of likely enemy encounter, the en-tering team received more detailed information fromthe robot operator. If interference was not expected,the squad or platoon leaders were satisfied to knowthat nothing suspicious had been observed. Thus, itis initially of more interest to identify threats such asenemies or IEDs than receiving a spatial outline. Thesoldiers did not always follow the path of the robot,both for physical and tactical reasons. For example,the robot could be used to enter through openingstoo narrow for a person; or, if moving between sto-ries, the robot could be lowered with rope or useanother set of stairs. When the platoon decided toadvance in another direction, the robot could beused for continued exploration of the excluded re-gion. According to MOUT-doctrine, a platoon onlyexpands one direction at a time to avoid fratricideand maintain a focused strike force; the robot thusallowed for a change in doctrine.

If given enough time, the operators were able tomake fairly accurate observations of the areas ex-plored with the robot �Figure 9�. Areas indicated butnot open to exploration, such as rooms behindclosed doors, however, tended to be neglected andforgotten. A separate experiment showed robot map-ping, on average, to require 70% more time and re-sult in 44% more errors compared to manual map-ping �Lundberg & Christensen, 2006�. From a tacticalaspect the approximate layout and character of thepremise is of main interest; dimensions do not needto be highly accurate. Other information of interestinclude appropriate positions for cover and strategicfire, influenced, among other things, by the material

and thickness of walls, doors, windows, and otherobjects in the area. The true effects of ammunitionand scatter are diminished during training withblank ammunition. Soldiers tend to seek cover be-hind objects which would in reality not withstandsmall arms fire and shrapnel. Improvements in theDirect Fire Weapon Effects Simulator system is cur-rently under implementation �by equipping build-ings with sensors and actuators� in order to modelthis aspect with more accuracy. Training facilities arealso being equipped with video and position track-ing devices for soldiers and vehicles. It is importantto also integrate these systems on robots as the user,at least in this case, spends more time practicingthan performing live missions.

Hostile encounter probably means loss of the ro-bot in return for decreasing personal risk. The sol-diers not only concluded that the robot could de-crease danger, but also argued that it could be usedas an alternative to hand-grenades and rifle fire dur-ing a cleanse operation—an ability that would de-crease the amount of required ammunition andminimize accidental harm to civilians and infrastruc-ture. One of the platoon leaders pointed out that therobot could spare enemy lives by opening for inter-

Figure 9. The operators were able to make fairly accuratesketches of areas they explored with the robot, in this casea two-bedroom apartment. While it took the robot opera-tor 18 minutes to do this sketch, it would only take thesoldiers a couple of minutes to perform the searchmanually.






B-07-0014_20214] action in situations traditionally handled with bruteforce. Voice communication through the robotwould enable for alternative solutions to hostagesituations, threatening crowds, and assistingwounded personnel.

The tactical significance of the robot dramati-cally changed when given lethal ability; turning itinto a tool for breaking up entrenched situations ordirecting action against enemies encountered duringexploration. Making the robot lethal also added thepossibility to use it for deterrent, threatening, or mis-leading purposes. Correct identification of friends,foes, and civilians is a crucial aspect during weapondeployment. This was a demand that could not bemet with the quality of video feedback providedfrom the Packbot Scout. The Claymore mine could,therefore, only be deployed in areas certain to holdsolely enemies.

The flashlight-payload was used to evaluate thebenefit of personnel not having to hold the lightthemselves. The system was deployed on a few oc-casions, but never during enemy encounter. TheUGV was not deployed to trigger anti-personnelmines or trip-wired devices as this threat was notincluded in the maneuvers. Trials were, however,made to visually identify IEDs during individual op-erator training. IEDs on floor level could occasion-ally be identified from within a few meters. Higherplaced objects were harder for the operator to spot.

4.2.5. Drawbacks

The test disclosed two main perceived risks, namelythat the robot might create delays in situations whentiming is crucial and that it could reveal the unit toenemies22 �3 and 1, respectively, in Figure 10�. It wasalso considered possible that having access to a ro-bot could make the soldiers reluctant to put them-selves at risk and thereby decrease the performanceof the unit �6 in Figure 10�. Throughout the trials thecommanders generally chose not to deploy the robotif they felt a rapid action would be beneficial; how-

ever, during the interviews they stated that the pri-oritizing of tactical benefits on behalf of increasedrisk was probably not feasible to such an extent dur-ing real missions. The time constraints are funda-mentally different in MOUT than compared to, forexample, EOD. EOD-robot missions are to a largeextent performed by a small specialized team, re-quested to come and act independently on a pre-specified and static target within an area seized anddefended by another unit. While EOD-robots areconsidered to decrease time-on-target, the MOUT-robot will not reduce the operational time �Lundberget al., 2007a�.

22Value the disadvantages the robot might convey. Estimate both theprobability of the event and the consequence to the unit deploying therobot…, 1. Robot operations reveal own unit, 2. Robot exposes operatorto increased risk compared to other soldiers, 3. Using the robot delays theunit during high ambition missions, 4. Using the robot delays the unitduring low ambition missions, 5. A plan has to be changed because therobot fails, 6. The availability of the robot makes the soldiers less willingto take risks and thereby decreases the capacity of the unit. �Replieswere given on a five point scale.�

Figure 10. The ratings of probability and consequencegiven for six disadvantages from robot use �see footnote22�. A combination of probability and risk shows that al-ternative 1 and 3 are considered the most critical. Alterna-tive 6 was also rated a serious risk. Dots indicate meanvalue and whiskers show the interquartile range.






B-07-0014_20214]

4.2.6. Benefits

Multi-role capabilities are desired to enable frequentdeployment. This in order to justify the added work-load the robot entails to the MOUT personnel. Thecomplexity of payloads spans from simple solutionssuch as two-way audio or fire circuits,23 to payloadsrequiring real-time data processing, or high speeddata transmission to the user-interface. The trial pe-riod did not enable a complete development, train-ing, and evaluation of the targeted scenarios. Therespondents were therefore asked to estimate thebenefit of the tested scenarios given that full imple-mentation would have been performed. 85% be-lieved that the system could be deployed 7–10 timesor more during a company attack of the type per-formed during training.24 None of the respondentsexpected the robot to be used less than four timesper attack.25

Figure 11 displays the estimated benefits from a

UGV in the tested missions.26 Weapons deployment,together with reconnaissance, were the applicationsregarded to be of most benefit. None of the respon-dents reported a strong opinion against using robotsas weapons.27 On the contrary, 79% replied that theyhad no objections to doing so; none of the respon-dents replied being strictly against. Ten out of the 41respondents did, however, comment that their opin-ion only applied as long as the fire command wasexecuted by a human �not autonomous/automatic�.

The tests were performed on maneuvers intensi-fied both in risk and time. To explore the offset, thetest participants were asked to value the benefits fordifferent types of operations likely for military forcesof the European Union �Ortega, 2005�.28 Of the fourtypes, the attended training maneuvers mostly re-semble the first. The first, together with the second,were also the ones believed to be most suitable forUGV deployment �Figure 12�. The final question ofthe questionnaire asked for general comments.Eleven people replied. Seven of these commented onthe importance of tactical development and trainingof UGV-behaviors; five were positive remarks aboutthe UGV.

4.2.7. Over-all Valuation

The users’ impression of the UGV’s overall costs andbenefits was investigated by asking them to comparethe value of the robot to two other systems also be-ing tested by the unit–inter-squad radios29 andnight-vision goggles.30 The robot was rated to be asvaluable as the night-vision goggles �mean 2.9 onthe five point scale�, and slightly less valuable than

23A fire circuit is a standard feature on EOD-robots that enablesthe operator to control a binary switch on the robot. The switch isnormally used to fire the disruptor gun, but it might as well beused to turn on/off any other payload mounted on the robot.24Rate how many times per company attack the UGV-system could havebeen deployed if each platoon had one system in use. � 0–3 times for theentire company, � 4–6 times for the entire company, � 7–10 times forthe entire company, � 11–15 times for the entire company, � 16 +times for the entire company. �Regularly two company attacks wereperformed per day during the maneuvers.�25The outcome was 0%, 15%, 44%, 20%, and 22%.

26Value what benefit the robot you have been testing would have in thefollowing missions if the tactics for this were fully developed andtrained…. 1. Reconnaissance and mapping of break-in point, 2. Recon-naissance and mapping of rooms or corridors, 3. Deployment of weaponssuch as Claymore mines, 4. Surprise or threat, 5. Illumination. �Thequestion handled 13 scenarios but only the 5 category-1 alterna-tives are included here �Section 1. Introduction�. Replies weregiven on a five point scale.�27Do you think that using weaponized robots is unethical?28Rate the value of the UGV for the following types of internationaloperations…: 1. Separation of parties by force, crisis management, peaceenforcing operations, 2. Conflict prevention, disarming, and confiscationoperations, 3. Evacuation operation of combatants and civilians, 4. Hu-manitarian assistance, catastrophe support, and evacuation of refugees.�Replies were given on a five point scale.�29How do you value the benefits of the UGV compared with the benefitsof the Inter-Squad Radio? �Reply was given on a five point scale.�30How do you value the benefits of the UGV towards the benefits of theNight-Vision Goggles? �Reply was given on a five point scale.�

Figure 11. Estimated benefits for scenarios tested �seefootnote 26�. Dots indicate mean value and whiskers showthe interquartile range.






B-07-0014_20214]

inter-squad radios �mean 3.5�. However, the re-sponses were rather dispersed �standard deviationof 1.3 for both�. The end-users were also asked tosuggest other equipment more important than aUGV. Ten people responded but none of the sugges-tions were reported by more than two people. Amore distinct reply was given to the question ofwhether or not UGVs should be acquired for theSwedish Rapid Deployment Force going into servicein 2008 �Nordic Battlegroup-08�.31 76% replied fullsupport for acquisition, 2% opposed strongly.

4.2.8. Tactical Findings in Summary

The most common mission was reconnaissance in-side buildings with uncertain enemy presence, andminor time constraints. The main benefits of the sys-tem were risk reduction and decreased weapon de-ployment. Major tactical drawbacks were reducedpace and increased risk of detection by the enemy.

Deploying the Packbot in MOUT is a two-person task due to both physical and mental de-mands. The operator and the robot system need tobe close to the front line to allow for rapid deploy-ment. Size, weight, mobility, robustness, and endur-ance live up to the demands. Night vision capabili-ties are necessary since indoor illumination cannotbe guaranteed. The users were of the opinion that

units given high-ambition tasks should be equippedwith one UGV per platoon. Additional sensors andpayloads would increase deployment rate. Arma-ment is desired and tactically significant. Table IIfurther summarizes the main tactical findings.

4.3. Technical Findings

The questionnaire investigated technical issues outof two perspectives: the first included technicalconstraints,32 the second the value of improve-ments.33 The second included both properties whichwere assessed during the trials and such that werenot �Excluded topics: robot armor, sensors, EOD-functions, autonomy�.

Narrow field of view, poor image quality, andlimited radio range were considered the most limit-ing features �Figure 13�. Robot speed also stood out.The robot’s top speed �4 m/s� was, however, ob-served to be limiting only in open spaces or out-doors. In most indoor settings the robot was able togo faster than the operator could control.

Automatic mapping �during teleoperation�, two-way audio, and the possibility to crop images fromthe video feedback for later viewing, were the high-est ranked improvements �Figure 14�. In addition tothe findings from the questionnaire, more descrip-tive information on technical performance was pro-vided through interviews and observations.

4.3.1. Robustness, Packing, and Transport

Packing and ruggedness are the primary featuresthat influence problems connected to bringing tech-nical equipment into the field. Equipment sensitiveto physical damage risks either breaking or not be-ing used due to the negative impact on the opera-

31Do you think UGVs should be included in Nordic Battlegroup-08…?�Reply was given on a five point scale.�

32To what extent do you regard the following properties to constrain theperformance of the tested robot? 1. Speed, 2. Off-road ability, 3. Radiorange/range of radio, 4. Low camera position, 5. Only having forwardlooking camera, 6. Quality of video feedback, 7. Power endurance, 8.Physical robustness, 9. Reliability, 10. Information transfer from therobot system to the persons needing the knowledge. �Replies weregiven on a five point scale.�33Rate the benefit of improving the robot in the following areas…: 1.Weapons deployment, 2. Automatic mapping, 3. Two-way audio, 4.Having several OCUs within the squad, 5. Improved OCU �smaller,lighter, easier to use�, 6. Acquire images, 7. Brighter LCD for use inbright daylight, 8. Decreased latency in video feedback and steer com-mands. �The question handled 17 improvements but only the 8category-1 alternatives are included here �Section 1. Introduction�.Replies were given on a five point scale.�

Figure 12. The value of the UGV in different types ofoperations �see footnote 28�. Dots indicate mean value andwhiskers show the interquartile range.






B-07-0014_20214]

Figure 13. The ratings of the level of constraint differenttechnical features imposed on the UGV’s performance �seefootnote 32�. Dots indicate mean value and whiskers showthe interquartile range.

Figure 14. The benefit of technical improvement of fea-tures that were available for trial �see footnote 33�. Dotsindicate mean value and whiskers show the interquartilerange.

Table II. Tactical findings in summary.


Tactical feasibility of size and weight Acceptable, although any decrease would be greatly beneficial. Inaddition to man-portability issues, fitting the robot into the vehiclesis a limiting factor.

Robot handling One operator and one assistantDurability of information renderedwith the robot

The informational value only lasts for a few minutes. As soon as thesoldiers enter the area, they will gain bettersituational awareness.

Command and stationing Positioning must allow for immediate access to the platoon commanders.Number of robots in a MOUT-company One for each platoon.Prerequisites Commanders on all levels must have knowledge about appropriate

situations for use, time allowances, and expected quality of information.Main disadvantages imposed by the robot The robot may delay advance, reveal the own unit, and make soldiers

reluctant to take risks of their own.Primary application duringtraining

Reconnaissance indoors, investigation of break-in point, and mapping.

Deployment frequency during training The robot was deployed 2–10 times a day. Missions were generallyshort, with a specified objective in mind.

Deployment frequency once fullyimplemented

Estimated to 7–10 times a day when performing offensive attack.

Applications with the most potential Reconnaissance, mapping, and weapons deployment.Benefits of the robot Decreased weapons deployment and reduced risk for own personnel,

civilians, and also enemy personnel. The ability to provide an alternativeto aggressive actions such as cleanse.

Relevant types of operation Offensive operations such as separation of contending parties, or peaceenforcement.

Outcome of enemy encounter The robot is easy to spot and eliminate, and enemy encounter willlikely lead to loss of the robot.

Lethal abilities A weaponed robot has a large tactical influence. Reliable identificationrequires improved situational awareness. The military revealed fewmoral considerations regarding use of weaponed robots.

Comparison to other equipment The robot was considered as valuable as night-vision goggles but lessvaluable than inter-squad radios.

Acquisition Systems such as the Packbot ought to be acquired for internationalmissions.






B-07-0014_20214] tor’s ability to act in unison with others. Ruggednessneeds to correspond to other equipment in use suchas radios, weapons, or binoculars. The heavier orbulkier an object is, the more physical damage it hasto withstand. Adequate carrying systems are espe-cially important for heavy equipment. The military’suniforms and carrying systems are adapted to exist-ing gear and cannot be expected to handle newitems such as a robot or OCU without modification.Packing is needed not only for the robot and theOCU, but for all system components such as charg-ers, spare parts, and batteries. During one of thefield studies, the spare batteries were found frozenin ice on the bottom of a 20-foot supply container;not an unlikely treatment of soldier gear, caused byharsh circumstances and time pressure in combina-tion with fatigue, carelessness, or lack of knowledge.

The Packbot Scout managed the realistic stresslevels of the trials fairly well. The camera covers forantennas, and the GPS receiver suffered the mostdamage, but connections and battery holders wouldalso benefit from additional robustness. The flippers34

are the most sensitive part of the propulsion systemand risked damage if the robot rolled down a stair-case or fell from a height over one meter. The Pack-bot was never submerged in water but it withstoodrain and splashes from driving through puddles.Subzero temperatures �Celsius�, aside from decreas-ing the battery performance, did not cause any prob-lems. The army is accustomed to performing theirabilities, with some variance in efficiency, regardlessof weather conditions, time of day or year; equip-ment needs to withstand the same conditions, orthey risk being viewed as inadequate. This stands incontrast to both manned and unmanned aviation forwhich weather limitations are tolerated.

4.3.2. Off-Road Ability and Audio

MOUT-soldiers are almost certain to encounter ob-structions such as misplaced or broken furniture ordemolished buildings. A robot with an inability totraverse obstructions such as steps or stairs will limitthe number of possible applications to such an ex-tent that the system is unsuited for military applica-tions. The environment of MOUT will be more de-manding than for service robotics is used, but notnecessarily as tough as for search and rescue appli-

cations �Carlson & Murphy, 2005�. During the test,ground conditions such as high grass, sand, andsnow challenged the skid steering of the Packbot,but only on very rare occasions did the engine over-heat or the tracks dislocate. Snow created the worstconditions for driving since snow gathered betweenthe tracks and the driving wheels, which increasedthe track tension. In soft snow the Packbot easily gotstuck because of snow piling up under its chassis. Insome cases the robot could come unstuck by drivingwith the flippers folded down. However, snow hadgenerally by then piled up on the ramp in front ofthe camera and blocked the view. The Packbot wasnot feasible for operation in more than five centime-ters of soft snow.

The flippers, which are a key component of therobot’s terrain ability, can be used to recover fromroll-overs. They also enable the robot to pass stairs,steps, and other obstacles. Ascending stairs is easierthan descending as the center of gravity in the fronttends to make the robot slide when going down.Barbwire or other thin metal wire snags in the tracksand ultimately traps the Packbot. The friction in thetransmission, which is normally high enough tokeep the robot still while no steering commands aregiven, is unable to prevent the robot from movingdown steep angles involuntarily. Such unintentionalmovement severely complicates driving in steep ter-rain �Lundberg & Christensen, 2007�.

The soldiers, however, generally found the Pack-bot’s capacity to pass obstacles to be sufficient�McBride, Longoria & Krotov, 2003�. In many casesmobility was not limited by the properties of the ro-bot but by the operator’s inability to grasp the set-ting through the user interface. If stuck, the robotcould often be retrieved as the operator could ob-serve the situation in line-of-sight. Noise from therobot, such as from collisions or slipping tracks,helped the operator understand the on-goings. Mul-timodal feedback �visual, audio, and other such astactile feedback of motor loads� would likely in-crease the operators driving performance signifi-cantly. The troops have hearing protection that elec-tronically blocks loud noise. The headset can receivein-line audio from radios or other electronic systemsand, thereby, enable the use of audio feedback evenunder noisy circumstances or stealth operations.Two-way audio is a standard feature on many EOD-robots which would be of tactical value also in

34Dual tracked arms in the front which are part of the robot’spropulsion system �Figure 3�.






B-07-0014_20214] MOUT. The robot was considered to be too noisy forstealth operation. In addition, booting the OCU op-erating system causes it to beep.

4.3.3. Operator Control Unit

In comparison to the robot, the off-the-shelf laptopwith external joystick did not suit the application aswell. The weakest point of the entire robot systemwas the USB-connection for the joystick. It was nomore rugged than that found on a common laptopcomputer, which became especially critical since areboot was required to regain contact with the joy-stick if the cable was momentarily disconnected.Both the USB port and the joystick were retrofittedwith protective covers along with a carrying systemallowing the operator to use both hands for climb-ing, crawling, and weapon deployment. The laptopserved as a basis during the trials but does not reachthe level of currently available portable or wearabletechnology �Hedström, Christensen & Lundberg,2006�. Major issues concern ruggedness, portability,daylight capacity of the LCD screen, and backlightfor the control buttons to allow identification indarkness.

4.3.4. Power Supply

On a full battery set, the system could, in general,cover half a day’s combat training.35 Current Pack-bot design requires a Philips screwdriver for batterychange; replacing the laptop batteries requires acoin.36 Charging was done by the only 230 V ACsource of the company, a gasoline-powered genera-tor by the staff tent. The generator was not run con-tinuously, but only when time allowed the staff tentto be put up. The generator went down every oncein a while, causing the Packbot battery charger toenter drain mode if the power was lost and regainedwith the battery inserted. Only one battery chargerper robot/laptop made charging time consuming.Lighter batteries or energy sources �instead of NiCd�would be beneficial since they constitute a large partof the system’s weight. Extra batteries could not bebrought along by the operators due to issues of

weight. Attempts to keep extra batteries in the com-bat vehicles often failed since the vehicles tended toreposition continuously. The medical evacuationpoint, which should always be strategically centeredand accessible to all units, proved to be a more suit-able place for spare batteries and modular payloads.The advantage of having the robot and operator con-trol unit run on military standard batteries would beimmense. Seven batteries out of twenty stoppedworking and two were mechanically broken duringthe two years of trials.

4.3.5. Radio Communication

The capacity and range of the radio link greatly af-fects the usefulness of the robot system. The practi-cal use of the system vanishes as the video framerate drops below ten frames per second. The IEEE802.11b link generally enabled the users to exploreup to two hundred meters outdoors and up to tworooms away inside of buildings. Unfortunately, thedistance between squad members was sometimesgreater than the range of the robot’s radio link. Thisprevented the operator from moving around withinthe range of his group, for example in order to briefthe squad leader, while maintaining contact with therobot. The circumstances for robot deployment out-side differ from inside. Outside, the distance be-tween safe spots are greater which makes the limitedrange even more restraining than indoors. The rangeof two hundred meters was often too short for out-door missions. Other radio-traffic on the 2.4 GHzfrequency notably decreased the quality of the Pack-bot’s radio link. In addition to a wider range, flexiblesolutions for dealing with jamming, national spec-trum regulations, and other radio traffic are desired.Integration with the user’s ordinary radio- andcommand-and-control system needs to be consid-ered. A doubled radio range would increase the tac-tical performance in MOUT notably.

4.3.6. Sensors

Area surveillance is one of the most common mili-tary tasks and was therefore tested with the robot.Unfortunately, the current interface design made itimpossible to perform in practice. To do visual sur-veillance using the 240�320 pixel image was simplytoo unstimulating a task for a human to performover time. Acoustic feedback would have increasedthe operators chance to regain attention when

35The OCU-laptop was delivered with one battery and a floppydisk drive. Replacing the floppy drive, which was of no use, withanother battery made the laptop and robot battery times corre-spond better to each other.36It is standard to avoid the need of tools for basic operation ofmilitary equipment.






B-07-0014_20214] needed, but the obvious solution would be amotion-detection functionality. The limited resolu-tion and field of view, in general, gave the operator afair chance to detect human-sized targets in smallrooms, such as apartments, and targets of car-size inlarger rooms such as industrial buildings. The visualfeedback did not enable a reliable distinction be-tween friends and foes �for example, by distinguish-ing uniforms�. The low placed cameras were easilyblocked by obstacles. The low placement also madeit hard for the operator to discover negative ob-stacles such as a staircase approached from above oredges of balconies. This became very clear in the in-dustrial facilities, where handrails and fences are de-signed for adults only and, therefore, regularly leavethe space closest to the ground unprotected. Basicknowledge about the robot’s design makes it easy toanticipate and avoid its field of view. With the cur-rent sensor setup, the robot is more suited to gainspatial information rather than find moving targets.

The IR-camera �close to visible spectrum� wasthe more used of the two cameras onboard. The fish-eye camera would be very useful as it gives abroader view which makes passing narrows easier;unfortunately, it requires normal indoor light. Thepoor light conditions prevailing during MOUT-operations made the IR-illuminator an importantfeature. Direct light, such as strong or low angledsunlight, blinds out both onboard cameras when, forexample, in a dark room facing a bright opening; theIR-illuminator can to some extent compensate forglare in close range. It should be noted that the IR-illuminator, which is invisible to the human eye, canbe seen through both the night vision goggles andthe night vision mode of commercial camcorders.The Packbot renders a thermal profile that is as vis-ible as a human when observed through a IR sight.37

The GPS did not provide any useful data in-doors or near buildings. In open terrain the range ofthe radio link prevented the robot to operate from adistance, to make GPS-positioning a relevant feature.

4.3.7. System Perspective

From the users’ perspective, a UGV system includesmore than a robot, batteries, and an OCU; militaryout in the field are accustomed to being providedwith all the bits and pieces required along the equip-

ment lifecycle. This includes technical manuals intheir native language, tactical guidelines, transportcasing, special tools, spare parts, training courses,etc. Achieving this will require joint efforts from in-dustry, third part developers, and users.

Successful integration of UGVs in the armedforces requires consideration of a broad range of is-sues, such as power supply, radio communication,maintenance, and command-and-control. The ongo-ing process of equipping each soldier with a wear-able computer, sensors, and radio has to be consid-ered in the design of future OCUs. A wide range ofstandards must be complied with. Some examplesare: Environment–MIL-STD 810, System Safety–MIL-STD 882, Human Factors–MIL-STD 1472F, Certifica-tion for Deployment in Explosive Atmospheres–ATEX Directive 94/9/EC, Nonmagnetism–Stanag 2897Annex C,38 and Electromagnetic Interference: MIL-STD 461/462

4.4. Technical Findings in Summary

Increased visual feedback, longer radio range, andtwo-way audio are the most desired improvements.GUI features such as snap-shot and automatic map-ping are valuable. Operations outdoors are re-strained by radio range, robot speed, camera resolu-tion, and glare. The robot operates more efficientlyindoors. Transmission noise restrains advancementby stealth. Integration with other military equipmentsuch as radios, command-and-control systems, andenergy supplies largely influences future overall per-formance. Table III further summarizes the maintechnical findings.

5. DISCUSSION

5.1. Changing the Doctrine

The prevailing MOUT-doctrine has evolved throughsmall iterations under a long period of time. Themain tools in use such as assault rifles, grenades,armored combat vehicles, radios, mines, and anti-tank weapons have all been around for 50 years ormore. EOD and mine-clearance robots are also usedon a regular basis to solve well-defined tasks accord-ing to highly-developed methods. Robots in otherapplications, on the other hand, constitute an en-

37Such as the one on the man-portable anti-tank missile BoforsMissiles Robot 56 BILL. 38Nonmagnetism is a requirement for EOD tools.






B-07-0014_20214]

tirely new functionality which sets aside many of theconstraints that have shaped prevailing tactics. Thisproject has shown that it is possible to change thewell-established and highly-structured doctrines toincorporate new concepts, but that it will require alarger tactical development effort than what is nor-mally accomplished with the users’ development as-sets. The incorporation of new technology can be di-vided into two categories:

1. Tasks performed today for which robotscould replace humans. This category, for ex-ample, includes the demanding and oftentime-consuming clearing IEDs or mines. TheEOD-teams typically first try to use the robotin order to avoid personal risk. If this doesnot succeed, they will proceed to solve thesituation manually under the protection of abomb suite.

2. Tasks that are not performed today but thatcould be accomplished using robots. Again,IEDs and mines can be used as an example.According to current strategy, MOUT-troopsdo not attempt to disarm or pass encounteredIEDs or minefields. Instead they try to find analternative route or make a request for anEOD-team to clear a passage. Equipping alsocommon units with EOD-robots could enablea complete change of some aspects in the cur-rent doctrine.

During the project it appeared that scenariosfalling into the first category were more concreteand, therefore, best suited for initial implementation.

It meant taking an established behavior in currentuse and modifying it to include the robot. If the newbehavior were to fail, the traditional methods couldbe used for recovery. However, gaining insight intothe full potential of new technology by including thesecond category would require a substantial rede-sign of doctrines, including identification of nichesthat might not have been targeted before. The oppor-tunity to test previously nonexisting abilities mightnot occur during current test and training maneu-vers as they are typically arranged to suit existingmethods. Applications that fall into the second cat-egory would result in behaviors that lack traditionalalternatives, which would make the user wholly de-pendent on all-new features. During the process ofintroducing new abilities, it should be kept in mindthat traditional methods might be set aside, whichcalls reasoning about how to act if the new featurewere to fail.

5.2. How to Implement

Previous experiments and demos have shown thatsoldiers and lower level officers in general are scep-tical about robotics until they get to fully know thesystem abilities �Lundberg, Barck-Holst, Folkeson &Christensen, 2003; Lundberg, Christensen & Hed-ström, 2005�. The hand-over of the robot system tothe users rather than bringing it to each appointedtrial, intended to give them a sense of responsibilityand, thereby, increase their commitment in deploy-ment. Still, the study showed that implementation ofa robot required significant efforts in development

Table III. Technical findings in summary.


Limiting technical properties Narrow field of view, poor image quality, and limited radio range.Desired features Automatic mapping, two-way audio, and possibility to crop images.Ruggedness and robustness The robot managed fairly well, but the operator control unit needs to

be adapted to portable field use.Off road ability With the exception of soft snow conditions, the robot managed fairly

well off road.Power endurance Meet the demands well.Sensing The night vision capability proved to be very valuable,

The GPS did not serve any purpose.Stealth Noise was the most revealing factor; the robot has an IR-profile

comparable to that of a person.Integration with other systems Making use of military standard components such as batteries would

simplify handling and maintenance of the robot. There are a numberof procedural, environmental, and military standards to comply with.






B-07-0014_20214] and training of new behaviors. Specific support andcoaching will be required to enable this; simply pro-viding a user with hardware will not be sufficient toimplement robots in teamwork applications requir-ing either human-robot interaction as well asorganization-robot interaction.

In addition to the two conscripts, an officer wasalso trained in robot operation, simply because offic-ers are accustomed to master all the skills of the sol-diers. Having an officer trained denoted knowledgeon a high level about the robot system’s capacitiesduring tactical planning and briefing. The trainedcaptain was second company commander and,thereby, a key person in the company.

The participants in military maneuvers are con-stantly graded, and leaders who attempt to deploynew features, instead of using traditional methods,take an increased risk of failure. Establishing a toler-ant, supportive, and rewarding atmosphere duringfuture tests will increase the rate of deployment.

5.3. When to Test

Performing the tests in a relevant environment is ofmajor importance. In this case the most realistic set-tings available were the large-scale maneuvers per-formed during soldier training. It should, however,be kept in mind that training maneuvers differ fromreal missions regarding mission profile and risk will-ingness.

A qualitative approach seems to be the best op-tion while testing in a large and complex settingsuch as one where several hundred persons act indi-vidually and dynamically on a mutual task. Quanti-tative measures are hard to apply as the course ofevents cannot be controlled and as the sought occa-sion might not occur often enough to be valued bystatistics, but rather have to be regarded through theopinion of experienced participants. Further, it is notpossible to perform repeated trials in the same testenvironment with a single test unit because of thelearning effect.

Regardless of what methodology is applied dur-ing the phases of research and development, most ofthe army’s evaluation is done through participatoryobservation, i.e., an embedded researcher �the officersand soldiers� uses the tested gear on a daily basisand forms a personal subjective opinion. Hence,even if qualitative approaches might not be the mostscientifically distinct, they correspond to howfielded products will be regarded by the end users.

In order to provide well-founded viewpoints itis crucial for the test-user to have some level of hard-ware knowledge as a base for reasoning. An impor-tant quality of long-term testing is the decrease ofbias connected to the introduction of a new product.It also gives the test group time to modify their be-haviors to the new circumstances and to develop amature opinion. In the beginning of the test period,the views on the system’s capabilities and possibleapplications differed vastly between users. The ini-tial interviews with the two officers produced nu-merous suggestions of how the robot might be usedin urban warfare. Most of these proved unfeasibleduring later trials. Similarly, the unstructured inter-views made during the trial illustrated that many ofthe suggestions of how to deploy robots were unre-alistic. Not until the end of the deployment phase,when the final interviews were conducted, were theusers experienced enough to reflect over robot de-ployment with more unity.

The more developed the introduced system is,the more relevant are the rendered results likely tobe; in conflict there is a need to perform testing inearly stages of the product design. An early intro-duction also enables parallel development of thesystem and the tactics for use. If this includes chang-ing well-established doctrines this process might re-quire significant time. When considering when toperform user testing it also has to be regarded thatthe users incorporated in studies will inevitablyform an opinion about the tested system. Unsuccess-ful trials might have an overall negative impact,which can be very hard to recover from; hence, thepoint of time when to perform testing is a strategiccompromise. The Packbot Scout was found capableenough to serve as a basis for research concerningsearch for objects, exploration, mapping, and pay-load delivery. Issues regarding mobility, endurance,robustness, radio range, user interaction, tactics, or-ganization, and ethics concerning arming robotscould be successfully investigated. Upon request,the users had the ability to see past properties thatrestrained the system, for example, the bulky opera-tor laptop. On the other hand, implementation of thePackbot did not give the users the ability to discusstopics like autonomy or sensor data fusion with thesame level of accuracy. Nor did the end-users haveenough background knowledge to value the systemin economical terms. On the other hand, it seemedmore natural for them to value the benefits of therobot in comparison to other equipment.






B-07-0014_20214] The project gave an opportunity to gain insights

about the physical and mental recourses availableamongst the users. It also proved to build a coopera-tion framework between the participating organiza-tions and to serve as a suitable way to initiate theuse of robotics in the addressed fields. Man-portableplatforms were a suitable first step in the introduc-tion of robots amongst the military. Dealing withlarger platforms, potentially with autonomous func-tionality, will increase the demands for testing byorders of magnitudes due to safety regards, legisla-tion, and practical issues such as transport, towing,fueling, training, and field repairs.

After the two years of trials, the officers of thecompany expect the robot activity to continue in or-der to keep the gained experiences up to date. Notproceeding might be regarded as negative by thepersonnel after their having identified robots as atool with high potential.

5.4. Collecting Data

Manuals and instruction videos allowed access tobasic users work procedures and terminology. Asmight be the case for many vocations, the documen-tation mainly covered the basics. Observation andparticipation were important means for gaining aholistic view of the users’ work practice. Attendingthe briefings proved valuable to grasp the course ofmaneuvers. The opportunities for participatory ob-servation proved to be very valuable for insights onthe individual soldier’s situation, as well as for es-tablishing an informal setting for discussions.

The monitoring of a group as large as a com-pany brought about a number of practical issues.The target of observation had to be constantlyshifted in order to catch the overall situation. Thefrequent shifts together with the movements of theunits in turn caused logistical demands such asbringing clothing, safety gear, supplies, and batteriesfor one or several days; arranging transport alongwith the combat vehicle convoys; and managing ac-commodation.

Circumstances during field studies made docu-mentation difficult. Photography proved to be themost valuable out of note-taking, photography, andvideo. Taking notes in the field often proved unprac-tical while video imposed too high workload com-pared to the information gathered.

The unstructured interviews conducted duringthe maneuvers were important in getting to know

the users and their activities. The short moments ofconversations in the field, however, did not allow forreflections. The ten interviews at the end of theproject served both as a recollection of the per-formed missions, and as a survey of the opinionsabout the system. Follow-up questions were used toverify the validity of the responses, which made itpossible to evaluate in which areas the users had awell-grounded opinion. The respondents showed ahigh level of cooperation and willingness to shareknowledge. Using an interviewer who had not par-ticipated in the field study decreased the risk of bias.The interviewer’s little previous knowledge did notseem to cause friction with the respondents, but re-sulted instead in more detailed and descriptive re-sponses, which opened the possibility for additionaldiscoveries.

The questionnaire aimed to document the per-formed robot missions as well as to investigate thequestions from the final interviews on a higher num-ber of respondents. The questionnaire and the teninterviews were designed and performed in parallel.Both the final interviews and the questionnaire indi-cated which topics the users could answer with va-lidity. A pilot-test of the questionnaire would haverevealed this and enabled a more feasible survey de-sign.

5.5. Delimitation and Outlook

The approach to have the end users deploy thetested product in a realistic setting limited the scopeto technology that could be operated by privates andalso have a fair chance to hold out in harsh environ-ments. The authors’ aims to motivate continued useof robots has favored applications that could showimmediate success over applications which requiremore extensive reformation, even though the lattermight have had a greater long-term potential. In-creased radio range, new payloads, improved userinterfaces, and tactical developments certainly holdpotential for future progress; however, this articlehas not focused on speculations about technical im-provements to come. The findings from current tech-nology, though, are promising enough to justifyimplementation for a number of reasons. The Pack-bot Scout as a tool for MOUT imposes an acceptableload on the troops, while offering advantages forcurrent uses. Even if the applications might not bethe most commonly used in international missions,the benefits reaped from the Packbot proved signifi-






B-07-0014_20214] cant enough to justify its presence in the toolkit forurban operations. Regarding possible future applica-tions, a premature implementation of the robotwould allow for tactical adaptations parallel to fu-ture technical developments. Early deploymentwould also enable a valid dialog between end usersand developers.

6. CONCLUSION

The long-term approach in a realistic setting has en-abled investigation of technical, tactical, ethical, orga-nizational, and human-robot interaction issues out ofthe users’ perspective. Gathered data became moreuniform over time, indicating the validity of the endresults and their ability to serve as a foundation forfuture work. However, it should be remembered thattraining maneuvers target the most engaging taskrather than the most common in active duty missions.

The study has showed that the MOUT-units relyon precise and thoroughly trained actions that can beexecuted with high precision and a minimum of am-biguity. The importance of reliability increases withrisk and uncertainty. Time is often a critical issue, andmeans of communication are often sparse. AllMOUT-gear must be portable as many of the mis-sions are performed on foot; ruggedness and weightare important issues. Equipment must function re-gardless of weather or time of day.

Deploying the Packbot in MOUT is a two-persontask for both physical and mental reasons. The mostcommon mission was combat reconnaissance in build-ings when enemy presence was uncertain and timewas not critical. The main benefits were reduced risksfor own troops and civilians, as well as a decreasedweapon deployment. The users were of the opinionthat units assigned to high-ambition tasks in urbansettings should be equipped with one UGV per pla-toon.

The range of the radio link, the video feedback,and the design of the operator control unit were thefeatures constraining the system’s overall perfor-mance the most. Other properties of the system, suchas ruggedness, size, weight, terrain ability, and en-durance did, on the other hand, prove adequate to theapplication. Trials with payloads indicated that thesystem has potential for more frequent deployment ifextended with modular add-ons for enhanced sens-ing, voice communication, and weapons deployment.

Integration with other military equipment must be aconsideration for the future design of robots.

The question of beneficial deployment is, how-ever, beyond just technical functionality. The intro-duction of robotics as a tool to infantry soldiers maybe comparable to automatic rifles or portable radios.Implementing a device with such novel functionalitywill require tactical development beyond what is nor-mally accomplished. Those who begin implementa-tion now will not only gain the benefits of today’savailable systems, but also be able to perform tacticaldevelopment parallel to ongoing technical develop-ment, and thereby shorten the time to deployment ofthe next robot generation with years.

ACKNOWLEDGMENTS

The participation of the 6th Urban Warfare Com-pany of the Royal Life Guards as well as the finan-cial support from the Armed Forces is gratefully ac-knowledged.

REFERENCES

Ashley, J. �2006�. Fes update. Unmanned systems 24�2�,17–23.

Barnes, M., Everett, H., & Rudakevych, P. �2005�. Throw-bot: Design considerations for a man-portable throw-able robot. In SPIE Proc. 5804: Unmanned Ground Ve-hicle Technology VII, Orlando, FL.

Birchall, P. �1997�. The longest walk: The world of bombdisposal. London: Arms & Armour.

Carlson, J., & Murphy, R. �2005�. How ugvs physically failin the field. IEEE Trans Rob., 21�3�, 423–437.

Carroll, D., Nguyen, C., Everett, H., & Frederick, B. �2005�.Development and testing for physical security robots.In SPIE Proc. 5804: Unmanned Ground Vehicle Tech-nology VII, Orlando, FL.

Chamberlain, P., Doyle, H., & Jentz, T. �1993�. Encyclope-dia of German Tanks of World War Two. Arms & Ar-mour, London.

Ebert, K., & Stratton, B. �2005�. Supporting the joint warf-ighter by development, training and fielding of man-portable ugvs. In SPIE Proc. 5804: Unmanned GroundVehicle Technology VII, Orlando, FL.

Foster-Miller, I. �2007�. Talon robots. Retrieved February 52007 from: http://www.foster-miller.com/lemming.htm.

Gardner, C., Treado, P., Jochem, T., & Gilbert, G. �2006�.Demonstration of a robot-based raman spectroscopicdetector for the identification of cbe threat agents. In25th Army Science Conference, Orlando, FL.

Hedström, A., Christensen, H., & Lundberg, C. �2006�.






B-07-0014_20214] Springer Tracts in Advanced Robotics, chapter AWearable GUI for Field Robots, pages 367–376. Num-ber ISBN 978-3-540-33452-1 in 25, Springer, Berlin,Germany.

Hisanori, A. �2002�. Present status and problems of firefighting robots. In Proceedings of Society of Instru-ment and Control Engineers Annual Conference,Osaka, Japan.

Hoving, P. �2003�. Soldier modernization. Swedish Journalof Military Technology, Vol 2, 22–27, 2003.

Hüttenrauch, H., & Eklundh, K. �2002�. Fetch-and-carrywith cero: observations from a long-term user studywith a service robot. In IEEE International Workshopon Robot and Human Interactive Communication,Berlin, Germany.

iRobot �2007�. Packbot. Retrieved May 27, 2007, from:http://www.irobot.com/sp.cfm?pageid=109.

Jones, C., & Lenser, S. �2006�. Sentinel: An operator inter-face for the control of multiple semi-autonomousugvs. Retrieved January 06, 2007 from: http://www.auvsi.org/unmannedscience/newsletter/attachments/47/Jones_C.pdf.

Jones, H., Rock, S., Burns, D., & Morris, S. �2002�. Autono-mous, robots in swat application: Research, design,and operations challenges. In AUVSI InternationalConference on Unmanned Vehicles, Orlando, FL.

Krulak, C. �1999�. The strategic corporal: Leadership in thethree block war. Marine Corps Gazette, Quantico82�1�, 18–22.

Kumagai, J. �2002�. Techno cops-police robotic and elec-tronic technology. Spectrum, IEEE, Vol. 39�12�, 34–39.

Leger, P., Trebi-Ollennu, A., Wright, J., Maxwell, S., Bonitz,R., Biesiadecki, J., Hartman, F., Cooper, B., Baumgart-ner, E., & Maimone, M. �2005�. Mars exploration roversurface operations: driving spirit at gusev crater. InIEEE International Conference on Systems, Man, andCybernetics, Waikoloa, Hawaii.

Lundberg, C., Barck-Holst, C., Folkeson, J., & Christensen,H. �2003�. Pda interface for a field robot. In Proceed-ings IEEE/RSJ International Conference on IntelligentRobots and Systems, Las Vegas, NV.

Lundberg, C., & Christensen, H. �2006�. Evaluation ofmapping with a tele-operated robots with video feed-back. In IEEE International Workshop on Robot andHuman Interactive Communication.

Lundberg, C., & Christensen, H. �2007�. How to brake apackbot. In the video session of ACM/IEEE Interna-tional Conference on Human-Robot Interaction,Washington DC.

Lundberg, C., Christensen, H., & Hedström, A. �2005�. Theuse of robots in harsh and unstructured field applica-tions. In ROMAN: IEEE International Workshop onRobot and Human Ineteractive Communication.

Lundberg, C., Reinhold, R., & Christensen, H. �2007a�.Evaluation of robot deployment in live missions withthe military, police, and fire brigade. In Proceedings ofSPIE Defence & Security Symposium, Orlando, FL.

Lundberg, C., Reinhold, R., & Christensen, H. �2007b�. Re-sults from a long term study of portable field robot in

urban terrain. In Proceedings of SPIE Defence & Secu-rity Symposium, Orlando, FL.

Matsuno, F., & Tadokoro, S. �2004�. Rescue robots and sys-tems in Japan. In IEEE International Conference onRobotics and Biomimetics, Shenyang, China.

McBride, B., Longoria, R., & Krotov, E. �2003�. Measure-ment and prediction of the off-road mobility of smallrobotic ground vehicles. In Performance Metrics forIntelligent Systems, Gaithersburg, MD.

Murphy, R. �2004�. Human-robot interaction in rescue ro-botics. IEEE Systems, Man, and Cybernetics Part C:Applications and Reviews, special issue on Human-Robot Interaction .

Ortega, M. �2005�. Petersberg tasks, and missions for theeu military forces. Retrieved January 6, 2007, from:www.iss-eu.org/esdp/04-mo.pdf.

Sandberg, S. �2003�. Night vision and the human factor.Swedish Journal of Military Technology .

Schneider, F. �2007�. European land-robot trial �elrob�,fgan, Hammelburg, Germany. Retrived Febuary 1,2007 from: http://www.m-elrob.eu/.

Scholtz, J., Young, J., Drury, J., & Yanco, H. �2004�. Evalu-ation of human-robot interaction awareness in searchand rescue. In IEEE International Conference on Ro-botics and Automation, Orlando, New Orleans.

Silverman, D. �2006�. Interpreting Qualitative Data, page291. SAGE Publications, London, Great Britain.

Sion, L. �2006�. Too sweet and innocent for war?: Dutchpeacekeepers and the use of violence. Armed Forces& Society 32�3�, 454–474.

Smith-Detection �2007�. Smiths supplies lightweightchemical detector to advanced cbrn detection robot.Retrieved February 5, from: http://www.smithsdetection.com/PressRelease.asp?autonum=114.

SPAWAR �2007�. Man-portable robotic system. RetrievedMarch 7, 2007, from: http://www.nosc.mil/robots/land/mprs/mprs.html.

Thomas, P., & Macredie, R. �2002�. Introduction to the newusability. ACM Transactions on Computer-Human In-teraction.

Tomatis, N., Terrien, G., Piguet, R., Burnier, D., Bouabdal-lah, S., Arras, K., & Siegwart, R. �2003�. Designing asecure and robust mobile interacting robot for thelong term. In Proceedings of IEEE International Con-ference on Robotics and Automation, Taipei, Taiwan.

Wada, K., Shibata, T., Saito, T., Sakamoto, K., & Tanie, K.�2005�. Robot assisted activity at health service facilityfor the aged for 17 months: an interim report of long-term experiment. In IEEE Workshop on Advanced Ro-botics and its Social Impacts, Nagoya, Japan.

White, J., Harvey, H., & Farnstrom, K. �1987�. Testing ofmobile surveillance robot at a nuclear power plant. InIEEE Conference on Robotics and Automation, Ra-leigh, NC.

Yanco, H.A., & Drury, J.L. �2004�. Where am I? Acquiringsituation awareness using a remote robot platform. InIEEE Conference on Systems, Man and Cybernetics,Hauge, Netherlands.




Long-Term Study of a Portable Field Robot in Urban Terrainhic/Georgia-HomePage/Publishing...Wiley...

Documents

Transcript of Long-Term Study of a Portable Field Robot in Urban Terrainhic/Georgia-HomePage/Publishing...Wiley...