CS594, Section 30682:

Distributed Intelligence in Autonomous Robotics

Spring 2003Tuesday/Thursday 11:10 – 12:25

Instructor: Dr. Lynne E. Parker½ TA: William Duncan

http://www.cs.utk.edu/~parker/Courses/CS594-spring03

Outline

• Overview syllabus and class policies

• Overview of Distributed Robotics

• Introduction to topics to be studied this semester

Overview of Syllabus and Class Policies

(See handout)

Overview of Distributed Robotics

Research in Multi-Robot Systems: Growing Rapidly

How Rapidly is this Research Growing?

• To investigate, conducted an INSPEC* Search:– Yearly query, 1979 - 2001

– Searched for articles including at least one of the following terms:• Multi-robot• Multirobot• Cooperative robot• Collaborative robot• Distributed robot

* Citation index for physics, electronics, and computing

Results of INSPEC Search Show Enormous Growth in Multi-Robot Systems Research

# Arti

cles i

n IN

SPEC

0

50

100

150

200

250

300

350

1975 1980 1985 1990 1995 2000 2005

What are Reasons for Enormous Growth?

• Advances in individual autonomous robotics

• Advances in understanding of complex systems

• Increased computational capabilities

• Many potential application domains

• RoboCup influence

• Etc...

Many Potential Application Domains for Multi-Robot Teams

NASA: Space explorationSpace Exploration

Industry: Mining, construction, ...Mining

Military: Surveillanceand reconnaissanceSurveillance and Reconnaissance

U.S. Department of Energy: Hazardous site cleanupHazardous Waste Cleanup

Eight Primary Areas of Research in Distributed Robotics

1. Biological Inspirations2. Motion Coordination3. Communication4. Object Transport and Manipulation 5. Reconfigurable Robotics6. Architectures, Task Planning, and Control7. Localization, Mapping, and Exploration8. Learning

For each area: For each area: •• Different extents of studyDifferent extents of study•• Many excellent solutionsMany excellent solutions•• Open research issues remain in all areasOpen research issues remain in all areas

Relative Concentration in Each Area of Multi-Robot Systems

0

50

100

150

200

250

300

350

400

450

500

1 2 3 4 5 6 7 8Biological Inspirations

Motion planningCommunicationManipulationReconfigurable robots

Architectures, etc.

Localization, etc.

Learning

# Arti

cles i

n IN

SPEC

(Values based uponINSPEC search for years 1979 - 2001)

1. Biological Inspirations

Objective: Study biological systems to achieve engineering goals

• Communication– Auditory, chemical, tactile, visual, electrical– Direct, indirect, explicit, implicit

• Roles– Strict division vs. loose “assignments”

• Hierarchies– Absolute linear ordering, partial ordering, relative ordering– Purpose: reduction in fighting, efficiency

• Territoriality– Reduces fighting, disperses group, simplifies

interactions• Social facilitation/sympathetic induction

– Allows for efficient use of resources• Imitation

– Complex mechanism for learning

Leaf cutter ants

Bee colony

Example MovieSwarm-type dispersion: Andrew Howard, Univ. of Southern California

Biological Inspirations: Future Directions

HighHigh--impact, open research issues: impact, open research issues:

•• Can we actually demonstrate large numbers of physical Can we actually demonstrate large numbers of physical robots (>=100) working together?robots (>=100) working together?

•• How do we use autonomous, dynamic physical How do we use autonomous, dynamic physical interconnectivity (like insects) to enable collective interconnectivity (like insects) to enable collective navigation over challenging terrains?navigation over challenging terrains?

2. Motion Coordination

Objective: enable robots to navigate collaboratively to achieve spatial positioning goals

• Issues studied:– Multi-robot path planning– Traffic control– Formation generation– Formation keeping– Target tracking– Target search– Multi-robot docking

Parker

Murphy

Multi-Robot Motion Control: Keeping-Formation

Our research has led to new insights into achieving global control via local interactions: fundamental limitations in small team members achieving globally optimal solutions with limited global knowledge.

L. E. Parker, “Designing control laws for cooperative agent teams”, Proc. of IEEE International Conference on Robotics and Automation, 1993.

S. Carpin, L. E. Parker, “Cooperative Leader Following in a Distributed Multi-Robot System”, IEEE International Conference on Robotics and Automation, 2002.

Motion Coordination: Future Directions

HighHigh--impact, open research issue:impact, open research issue:

•• How do we generate robust motion coordination for 3D How do we generate robust motion coordination for 3D environments in previously unknown terrain amidst environments in previously unknown terrain amidst dynamically moving obstacles?dynamically moving obstacles?

3. Communication

Objective: Enable robots to exchange state and environmental information with a minimum bandwidth requirement

Issues studied:• Explicit vs. Implicit• Local vs. Global• Impact of bandwidth restrictions• “Awareness” • Variety of mediums: radio, IR, chemical scents,

“breadcrumbs”, etc.

Balch and Arkin

Jung and Zelinsky

Communication: Future Directions

High impact, open research issues:High impact, open research issues:

•• How do we enable multiHow do we enable multi--robot teams to work reliably in robot teams to work reliably in faulty communication environments?faulty communication environments?

•• Can we develop passive action recognition capabilities Can we develop passive action recognition capabilities in robot teams?in robot teams?

•• How do we enable multiHow do we enable multi--robot teams to set up robot teams to set up dynamic, mobile communications networks?dynamic, mobile communications networks?

4. Object Transport and Manipulation

Objective: Enabling multiple robots to collaboratively push, move, or carry objects that cannot be handled by one robot alone

• Issues studied:– Constrained vs. unconstrained motions– Two-robot teams versus "swarm"-type teams– Compliant vs. non-compliant grasping mechanisms– Cluttered vs. uncluttered environments– Global system models vs. distributed models– Etc.

Kube

Parker

Cooperative Baton Passing

• Our research has shown how mobile robots can perform tightly-coupled cooperation embedded in a “loosely-coupled” cooperative architecture.

Object Transport and Manipulation: Future Directions

HighHigh--impact, open research issue:impact, open research issue:

•• Can we achieve cooperative transport over uneven Can we achieve cooperative transport over uneven outdoor terrains?outdoor terrains?

5. Reconfigurable Robotics

Objective: Obtain function from shape, allowing modules to (re)connect to form shapes that achieve desired purpose

• Earliest research included reconfigurable/cellular robotics

• Several newer projects:– Various navigation configurations (rolling track,

spider, snake, etc.)– Lattices, matrices (for stair climbing, object support,

etc.)Castano et. al.

Movies of PolyBot(Mark Yim, Xerox PARC)

Tricycle pedalingStair Climbing

Porous material climbing

Reconfigurable Robotic: Future Directions

HighHigh--impact, open research issues:impact, open research issues:

•• How do we facilitate autonomous (vs. manual) How do we facilitate autonomous (vs. manual) reconfiguration?reconfiguration?

•• How can we extend concepts to practical applications?How can we extend concepts to practical applications?

6. Architectures, Task Planning, and Control

Objective: Development of overall control approach enabling robot teams to effectively accomplish given tasks

• Issues studied:– Action selection– Delegation of authority and control– Communication structure– Heterogeneity versus homogeneity of robots– Achieving coherence amidst local actions– Resolution of conflicts

Example 1: “Mock” hazardous waste cleanup -- Part I

L. E. Parker, “On the design of behavior-based multi-robot teams”, Advanced Robotics, 1996.L. E. Parker, “ALLIANCE: An architecture for fault tolerant multi-robot cooperation”,

IEEE Transactions on Robotics and Automation, 1998.L. E. Parker, “Evaluating success in autonomous multi-robot teams: Experiences from ALLIANCE architecture

implementations”, Journal of Theoretical and Experimental Artificial Intelligence, 2001.

Example 1: “Mock” hazardous waste cleanup -- Part II

Example 1: “Mock” hazardous waste cleanup -- Part III

Architectures, Task Planning, and Control: Future Directions

HighHigh--impact, open research issues:impact, open research issues:

•• Can a general architecture be developed that is easily Can a general architecture be developed that is easily tailored to fit a wide range of multitailored to fit a wide range of multi--robot systems and robot systems and applications? Or, are specialized architectures required?applications? Or, are specialized architectures required?

•• How do we quantitatively measure and compare different How do we quantitatively measure and compare different multimulti--robot team architectures?robot team architectures?

7. Localization, Mapping, and Exploration

Objective: Enable robot teams to cooperatively build models of their environment, or to accomplish spatial tasks requiring knowledge of other robot positions

• Issues studied:– Extension of single-robot mapping approach to multi-

robot teams– Hardware, algorithms for robot positioning– Sonar vs. laser vs. stereo imagery vs. fusion of

several sensors– Landmarks vs. scan-matching

Example Movie of Distributed Localization(Univ. of Southern California)

Localization, Mapping, and Exploration: Future Directions

HighHigh--impact, open research issues:impact, open research issues:

•• How do we reliably localize, map, and explore in 3D How do we reliably localize, map, and explore in 3D environments (especially with limited or unavailable DGPS)?environments (especially with limited or unavailable DGPS)?

•• As we extend to cooperative localization, how do we As we extend to cooperative localization, how do we determine the limit of effectiveness of adding more robots todetermine the limit of effectiveness of adding more robots tothe team?the team?

•• How do we merge information efficiently, in light of How do we merge information efficiently, in light of communication limitations, noisy sensors, and localization communication limitations, noisy sensors, and localization errors?errors?

8. Learning

Objective: Enable multi-robot teams to adapt or develop own control approach to solve a task with minimal human operator input

• Application domains studied:– Predator/prey– Box pushing – Foraging– Multi-robot soccer– Cooperative target observation

Parker

Example: Adaptive Box Pushing

L. E. Parker, “Adaptive Heterogeneous Multi-Robot Teams”, Neurocomputing, 1999.

Learning: Future Directions

HighHigh--impact, open research issues:impact, open research issues:

•• Can we generate efficient learning in inherently Can we generate efficient learning in inherently cooperative tasks?cooperative tasks?

•• How do we deal with credit assignment in groups How do we deal with credit assignment in groups ----determining individual robot contributions from group determining individual robot contributions from group performance?performance?

New Research Areas Receiving Increased Attention

• Robot-Agent-People teams as peers

• Heterogeneous teams

• Physical demonstrations of large numbers of robots (>=100)

Summary of Introduction to Topic

• Dramatic increase in research publications in multi-robot systems in last 5 years.

• Eight “traditional” topic areas studied within a variety of applications

• New research areas emerging

• In all research, before it will be used practically in real-world applications, need increased focus on:– Robustness– Reliability

Topics to be Studied this Semester

• Taxonomies, Metrics, Evaluation

• Biological Inspirations

• Low-level, homogeneous, swarm robots– Swarming, dispersion, homing, etc.– Search/coverage– Sensor networks– Communication and communications networks– Formations– Pursuit/herding– Tracking

• Reconfigurable robots

Topics to be Studied this Semester (con’t.)

• Higher-level strategies, Heterogeneous Robots– Multi-robot path planning, traffic management– Task allocation:

• Negotiation-based• Market-based• Modeling-based

– Marsupial teams– Air-ground teams– Multi-robot soccer

• Embedded Systems– Smart dust– Intelligent rooms/smart homes– Amorphous/pervasive computing

Preview of Next Class

• Biological Inspiration– Birds and Bees ;-)– Herds and schools– Stigmergy– Etc.

• Reading list distributed, presentation selections begin

• Homework 1 distributed