Optimal Wireless Access Points Placement in a Manufacturing...

Proceedings of the International Conference on Industrial Engineering and Operations Management Washington DC, USA, September 27-29, 2018

© IEOM Society International

Optimal Wireless Access Points Placement in a Manufacturing Facility using Genetic Algorithm

Tarek Ibrahim Hala A. Farouk Ahmed F. Abdel Moneim

Industrial and Management Engineering Department Arab Academy for Science and Technology

Abstract Streamlining materials ow in flexible manufacturing systems is the key factor to greatly reduce idle time costs and labor costs. It is achieved by building location-aware robots that can estimate its position using off-the-shelf hardware. The location-aware robot can determine its position inside a building using online measured Received Signal-Strength Intensity (RSSI) from different wireless Access Points (AP), and an offline database of previously measured RSSI. The accuracy of the calculated online position can greatly affect the preciseness of the material handling. This paper studies the impact of the location of the wireless access points and their quantity in a manufacturing facility on the localization error. An optimum number of wireless access points and their locations is determined using a Genetic Algorithm for the case study of the Egyptian textile manufacturing facility. The optimum wireless access point placement in the 17 500 m2 manufacturing facility improves the accuracy of localization by 46% using only nine APs to reach 4.7 meter error and 87% using 35 APs to reach 1.1 meter error, compared to the original plan of the wireless access points locations.

Keywords flexible manufacturing systems, autonomous mobile robots, indoor positioning, location fingerprinting, wireless nodes optimization

1. Introduction

Effective material handling in manufacturing system can reduce production cost by significant amounts since it accounts for 25% of all employees, 55% of all factory space, and 87% of the production time (Tompkins et al. 2003). Moreover, it is estimated that material handling represents 15% to 70% of the total cost of product manufacturing (Gamberi et al. 2009). Therefore, it is desired to implement automated material handling however with least cost and minimum changes in the manufacturing facility layout and equipment. There are two automated material handling solutions; Automatic Guided Vehicle Systems (AGVS) and Autonomous Mobile Robots (AMRs). However, AMRs are a more cost effective material handling solution compared to AGVS because AMRs have a lower cost of initial investment and can be recovered faster than in AGVS (Davich 2010). One of the fundamental problems of mobile robots is to locate their position in the warehouse or the manufacturing environment. Accurate localization has a direct impact on the quality, flow and running cost of the system. The AGVS uses a guidance system, tracks or floor markings, such as painted lines or glued- on reflective tapes. The permanent track is not desirable and constant wear on markings can cause system reliability issues. Since 1990, there is a large number of research in the field of indoor localization. Researchers have discussed different types of sensors, such as sonar (ultrasonic) sensors (Triggs & Cameron 1992); (Rencken 1993); (Li et al. 2016). Camera (vision) Sensor (Betke & Gurvits 1997); (Lee et al. 2015). Laser Range Finder (Sobreira et al. 2015); (Kim & Chung 2016), Radio Frequency Identifier (RFID) reader (Jing & Yang 2007), inertial sensors, wheel encoders (Skobeleva et al. 2016), and Wireless Sensor Network (WSN). Each type of sensor and approach has its advantages and disadvantages, however, Wireless Sensor Networks (WSN) utilize existing WiFi Access Points infrastructure present in every manufacturing system. Hence, they are low cost, require no physical guidelines and no additional hardware is needed for navigation. Locating the mobile robot using WSN is a low cost, low power consumption system with an acceptable accuracy (Chen et al. 2013). However, increasing the accuracy of positioning using WSN is a big challenge. In the meantime, there are two approaches to use the WiFi signal to locate the mobile robot; Trilateration and Fingerprinting approach. Trilateration approach is simple to implement, it geometrically calculates the robot position using the Received Signal Strength Intensity (RSSI) from three or more wireless access points. However, the wireless access points and the

2504



robot need to be in Line-of-Sight to obtain accurate localization results. In manufacturing systems, there are walls, doors, walking labour that may block the Line-of-Sight. Therefore, experiments show that the error of this approach can be up to five meters (Li et al. 2006). The fingerprinting approach is an effective method for the robot positioning (Kaemarungsi and Krishnamurthy 2004) ; (Nerguizian and Outemzabet 2008). The fingerprinting approach records the RSSI of each wireless access point for a large set of physical positions where the robot might be located in the future. The physical positions where the robot might be located are called reference points (RP). The RSSI is sensitive to noises, obstacles and moving objects and therefore can fluctuate especially in an environment such as a manufacturing facility and thereby increasing the error in the calculated robot location. Therefore, in a flexible manufacturing facility where the robot uses its position to locate jobs or handle materials, the localization error is encouraged to be lower than one meter. The location of Wi-Fi access points greatly affect the localization error. In manufacturing facilities, like any other facility, the location of the Wi-Fi access points is chosen based solely on coverage without taking into consideration that these Wi-Fi access points can also be used for localization and navigation (Alsmady and Awad 2017). Since the locations of the Wi-Fi access points play a significant role in the localization accuracy, this paper proposes a novel technique to find the most appropriate positions for the access points that maximize the variation of RSSI from one position to the other, which in turn maximizes the indoor localization accuracy. Several researchers studied the effect of changing the APs’ locations to achieve better indoor localization (Battiti et al. 2003); (Zhao, Yongxiang, Huaibei Zhou 2008); (Baala Oumaya Zheng 2009); (Liao et al. 2010); (Farkas et al. 2013); (Ismail et al. 2016). In (Zhao, Yongxiang, Huaibei Zhou 2008), authors develop an objective to find best configuration of APs by comparing a set of AP configurations using a selection function they have developed. In (Ismail et al. 2016), the selection function developed in (Zhao, Yongxiang, Huaibei Zhou 2008) is criticized and an alternative function has been developed and tested on similar configuration set to prove improvements. However, in both papers, the area under study is around 3 200 m2, which is considerably small. In (Alsmady and Awad 2017), authors develop an objective function to find the optimum location of the APs in a large area of around 36 000 m2 using Genetic Algorithm. However the layout and type of floorplan is not indicated and therefore the applicability of the model on manufacturing facility with many sources of noise and many obstructions is not confirmed. In this paper we design a Genetic Algorithm that searches the space of all RPs to find the optimum configuration resulting in minimum localization error on a 17 500 m2 manufacturing facility area. The floorplan of the manufacturing facility includes many obstacles which renders the search for optimum AP locations challenging. We implement the Genetic Algorithm using the selection function developed in (Ismail et al. 2016). Results show 46% reduction in localization error; from 8.7 meter to 4.7 meter, with the same number of APs (nine APs). The Genetic Algorithm developed in this paper uses a variable length chromosome, which is used to change not only the location of the APs but their quantity too. Results show that increasing the number of APs has also a great effect on the localization error dropping it to 1.14 meter at 35 APs and 1.1 meter at 89 APs, which is a further 76% reduction. We also analyze the rate of the localization error reduction and give recommendations for potential reductions of the localization error. The rest of the paper is structured as follows: Section 2 provides a background relevant to this work. Section 3 introduces the case study of the Egyptian textile manufacturing facility with its floorplan. Section 4 presents the developed Genetic Algorithm. Section 5 presents the experimental results. Finally, Section 6 highlights the most important conclusions drawn from this work and potential future works. 2. Related Work Indoor positioning systems using the fingerprinting approach are specifically designed to assist the robot to navigate in the manufacturing facility. The fingerprinting approach uses the indoor wireless signal propagation model to navigate as has been mentioned before. The wireless signal strength is dependent on many factors, including the physical characteristics of the warehouse or the manufacturing facility, the floorplan, the machines installed, etc. In this section the basic propagation model used is discussed. 2.1 Free Space Signal Propagation and Path Loss Model The basic signal propagation model could be derived from the inverse power law function. The power of the signal radiated from a source is dissipated proportional to 1

𝑑𝑑2 where d symbolizes the distance travelled, if we assume that

the transmitter and the receiver have an unobstructed line-of-sight between them. There are currently quite many experiments conducted to assess the signal levels at certain points in buildings. However, the free space power Pr of 2.4 GHz frequency signal propagation (Recommendation ITU-R P.1238-1 (2015)) received by an antenna which is separated from a radiating transmitting antenna with transmitted power Pt by a distance d, is given by Friis in Eq.1.

2505



𝑃𝑃𝑟𝑟(𝑑𝑑) = 𝑃𝑃𝑡𝑡𝐺𝐺𝑡𝑡𝐺𝐺𝑟𝑟𝜆𝜆

2

(4𝜋𝜋)2𝑑𝑑2𝐿𝐿(1)

where Gt and Gr are the transmitter and receiver gains, L is the system loss factor not related to propagation (L≥1) and 𝜆𝜆 is related to the carrier frequency. Pass loss PL is an indicator for the signal attenuation in decibel (dB) between transmitter and receiving antenna and therefore Eq.1 is given in logarithmic form in Eq.2.

𝑃𝑃𝑃𝑃(𝑑𝑑𝑑𝑑) = 10 𝑙𝑙𝑙𝑙𝑙𝑙10𝑃𝑃𝑡𝑡𝑃𝑃𝑟𝑟

= −10 𝑙𝑙𝑙𝑙𝑙𝑙10𝐺𝐺𝑡𝑡𝐺𝐺𝑟𝑟𝜆𝜆2

(4𝜋𝜋)2𝑑𝑑2 (2)

A convenient way to express the PL as a function of distance as all other parameters are constant for a specific environment is given in Eq.3.

𝑃𝑃𝑃𝑃(𝑑𝑑𝑑𝑑) = 𝑃𝑃𝑃𝑃(𝑑𝑑0) + 20 𝑙𝑙𝑙𝑙𝑙𝑙10(𝑑𝑑) (3) The reference distance d0 is determined by the measurements close to the transmitter. The mathematical expression taking into consideration the Multi Wall and Floor model (MWF) proposed by (Lott and Forkel (2001)) in addition to the European ITU-R P.1238 model from the (Recommendation ITU-R P.1238-1 (2015)) is given in Eq.4.

𝑃𝑃𝑃𝑃(𝑑𝑑𝑑𝑑) = 𝑃𝑃𝑃𝑃(𝑑𝑑0) + 10 𝑛𝑛 𝑙𝑙𝑙𝑙𝑙𝑙10(𝑑𝑑) + 𝑘𝑘𝑘𝑘 (4) The parameter n is called the Path Loss Exponent and it depends on the specific propagation environment conditions as in Table 1(Rappaport 2002), such as the type of the building construction, architecture and location of the point where the measurements are carried out. (n = 2) in the case of free-space path and (n > 2) in the case of presence of strong effects degrading the signal strength and k is the number of obstructions between the transmitter and the receiver, and F is the obstruction attenuation factor.

Table 1: Typical values of path loss exponent for different environments

Environment Path loss exponent Free space 2 Urban area and cellular radio 2.7 to 3.5 Shadowed urban cellular radio 3 to 5 In building line-of sight 1.6 to 1.8 Obstructed in building 4 to 6 Obstructed in factories 2 to 3

2.2 Fingerprinting Approach The fingerprinting approach has been accepted as an effective method for WiFi localization (Sun Yuxiang Liu 2014). It can be followed in a deterministic, probabilistic or machine learning based method (Dellaert et al. 1999; Li et al. 2006; Baggio and Langendoen 2008; Nerguizian and Outemzabet 2008); Jain et al. 2012). It stores only averages of signal strength values for each location obtained from training data and discards remaining information. In order to reach localization error less than one meter, the probabilistic nature of the signal strength has to be taken into consideration by either storing multiple measurements over time or by making many measurements enough to _t it into one of the known probability distribution functions (Navarro et al. 2010; Biswas and Veloso 2010; Le Dortz, N. and Gain, F. and Zetterberg 2012; Milioris et al. 2014). However, the size of the database required for the probabilistic fingerprinting method is greatly larger than the deterministic method. In general, in order to locate the robot there are two phases shown in Fig.1; the offline training phase and the online location determination phase (Li et al. 2006).

2506



Figure 1. Fingerprinting Steps

In the offline phase, the test area is decomposed into a grid. Each grid node is called Reference Point (RP). The location fingerprints are collected by performing a site-survey of the Received Signal Strength Intensity (RSSI) from multiple access points (AP) (Fig. 2). The RSSI is measured with enough statistics to create a database or a table of predetermined RSSI values at each RP. This table is called radio map. The vector of RSSI values at a grid point is called location fingerprint of that point. In the on-line phase, the robot send a vector of measured RSSI values from different APs to a central server. The server uses an algorithm to estimate the robot location and reports the calculated position back to the robot. The most common algorithm to estimate the robot location calculates the Euclidean distance between the measured RSSI vector and each fingerprint in the database (M. & A. 2002); (Baala Oumaya Zheng 2009); (Le Dortz et al. 2012). The coordinates associated to the fingerprint that provide the smallest Euclidean distance is returned as the estimated position.

In order to reduce the localization error various methods found in literature attempt to improve parts of the fingerprinting localization system shown in Fig.1. In (Biswas & Veloso 2010); (Mirowski et al. 2012); (Lee et al. 2015) authors attempt to improve the localization accuracy by adding an extra sensor in addition to the wireless access points, such as laser, vision or any other sensor mentioned earlier in Section 1. A new dimension to the input of the localization algorithm is thereby added. However, this approach increases the cost of the system since other sensors are expensive as explained before. Other research emphasize on the algorithms used for estimating the location, which associates the fingerprints with the location coordinates (M. & A. 2002); (Pandya et al. 2003); (Wang et al. 2003). The more complicated the algorithm becomes the more computational speed it requires. Other research considers the understanding of the statistical characteristics of the RSSI to be crucial in improving the localization error and therefore a training phase is required to develop a probability distribution for the RSSI. These methods are called probabilistic approach and they change the database contents in the offline phase (Le Dortz et al. 2012); (Chen et al. 2013); (Ma et al. 2015). Other research have focused on increasing the RPs to improve the localization error ( Li et al. 2006).

Figure 2. Example Reference Point (RP) Data

2.3 Robot Localization Algorithm The robot at a certain location sends a vector of RSSI values taken from n wireless access points. The localization algorithm uses this vector and compares it to the many vectors of the RPs saved in the database to find the closest one and accordingly estimates the robot location. Many algorithms can be used to estimate the position of the robot.

2507



However, the basic method is to use the Euclidean distance between the elements of the RSSI vector sent from the robot and the elements of each of the RP RSSI vectors stored in the database as shown in Eq.5 [(Li et al. 2006)]. The RP resulting in the smallest Euclidean distance is estimated to be the closest location to the robot's actual location. Usually, the estimated location and the actual location are not a perfectly matched and therefore, the Euclidean distance between the estimated and the actual location is called the localization error.

𝑑𝑑 = �∑ (𝑠𝑠𝑘𝑘 − 𝑆𝑆𝑘𝑘)2𝑛𝑛

𝑘𝑘=1 (5)

where 𝑠𝑠𝑘𝑘 is the 𝑘𝑘𝑡𝑡ℎ element in the vector of RSSI values sent from the robot to the server and 𝑆𝑆𝑘𝑘 is the 𝑘𝑘𝑡𝑡ℎ element in the vector of RSSI values of one of the RPs stored in the database. In order to reduce the localization error, the estimated robot location is not determined by a single RP with the smallest Euclidean distance, but by a group of neighboring RPs having smallest Euclidean distances. These methods are called the k Nearest Neighbor (kNN) or the Weighted k Nearest Neighbor (WkNN). As the name implies, in the kNN algorithm, a number of k neighboring RPs with smallest Euclidean distances are used to estimate the robot location. In the WkNN algorithm, also the k neighboring RPs with smallest Euclidean distances are used in the localization algorithm however with weights. 3. Problem Formulation A two-dimensional floor area of 162x108 [meter] similar to a manufacturing facility with a warehouse is considered. The RP locations are taken with one meter spacing, resulting in 17,496 RP. In ideal conditions, the localization error will be only one meter. In this manufacturing facility, Wi-Fi access points are placed optimally to reduce the localization error at robot location. The overloading of the manufacturing facility with WiFi APs is not a solution. Since, network management becomes more complicated and the increase in the electromagnetic field interference with other WiFi signals renders the Wi-Fi APs useless. Therefore, the manufacturing facility has been surveys for all candidate locations for the Wi-Fi access point which are secure and away from hazards. The survey has resulted in 80 candidate locations. The RSSI has been recorded at various locations in the manufacturing facility to determine the reference distance d0 required for the RSSI calculations using the propagation model in Section 2.1. It has been noted that the RSSI varies over time due to many factors mentioned before, such as machine noises, people or objects moving. The variation in the RSSI signal recorded at a single RP is shown in Fig.3. A histogram of the RSSI values at one of the RPs is shown in Fig.4. However, the variation are not completely random but have a mean value, which can be used.

Figure 3. RSSI signal strength varying over time for several APs

2508



Figure 4. Histogram of the Recorded RSSI Values at a Single Location in the Manufacturing Facility

To formulate the problem formally, we consider a set of M candidate locations (L1, L2,..... , LM) in which it is possible to deploy any type of access points. It is required to deploy N wireless access points into a set of M locations (N << M). The configuration Q is a set of chosen candidate locations (Q ϵ M) in which access points have been deployed. At each reference point (RP), signals received from each access point can be calculated using the propagation model in Section 2.1 and combined into a vector FPRP = (RSSI1, RSSI2,….., RSSIN). This vector FPRP is called the RP fingerprint. The objective of this problem is to place N as minimum as possible APs that minimizes the localization error without losing the signal coverage for the end user. 4. Optimization Algorithm A selection algorithm developed in [Ismail et al. (2016)] is used as the fitness function for the Genetic Algorithm developed in this paper. The algorithm takes number of APs and the Configuration Q as input and calculates the fitness (J) of this configuration. The algorithm includes another input P, which is the penalty value for configurations with dead zones. Configurations with dead zones are configurations in which areas are not covered by any wireless access point. The algorithm uses rmax = -17, which is the reference distance d0 of the propagation model in Section 2.1. It uses also rmin = -80 as the minimum RSSI value in dB at which communication is not considered working between the robot and the access point. The weights w1 and w2 have been adjusted to the floorplan under study as has been recommended by the authors in [Ismail et al. (2016)]; w1 = 0.8 and w2 = 0.2. Algorithm 1 Calculate the objective function J for a specific number of Access Points N for specific configuration Q Require: N is the Number of Access Points Require: P is the penalty for dead zones Require: Q is the Configuration, which consists of N Access Points (AP1, AP2…, APN) Ensure: J is the Objective Function Value for the configuration Q 1: function Objective Localization B (N, Q) 2: for All Access Points APs do 3: rtotal ← sum(rscaled ← (RSSI - rmin) / (rmax - rmin)) for all RPs 4: end for 5: raverage ← rtotal/N 6: for All Access Points APs do 7: vtotal ← sum( (rscaled - raverage) 2 ) for all RPs 8: end for 9: rvariance ← vtotal/N 10: J ← w1 * raverage + w2 * rvariance + P 11: end function The chromosome used in the developed Genetic Algorithm is of variable length to accommodate configurations with variable number of access points. Each gene in the chromosome is an integer value from 1 to 90, since there are 90 candidate locations for the access points. The gene value carries the index of the candidate location in which an access point will be deployed. The startup population consists of 200 parents. The crossover and the mutation are performed by the constraint dependent function. The crossover fraction is 0.8. Reproduction replaces the 10 least fitting parents

2509



with the 10 most _t parents. The Genetic Algorithm from the Global Optimization toolbox of the MATLAB computing environment has been used. The fitness function of each generation is plotted in Fig.5 showing the convergence of the Genetic Algorithm.

Figure 5. Fitness Value of the GA Optimization Tool

5. Experimental Results The algorithm computing the fitness function J shown in Section 4 has been used to study the impact of changing the locations of the wireless access points without changing the number of access points used. We have conducted the experiment using nine wireless access points and the resulting localization error is shown in Fig.6. Changing the location of a single wireless access point can drastically affect the localization error. The localization error has been reduced by 46%; dropping from 8.7 meter to 4.7 meters. A heatmap for the first configuration and the fourth configuration is shown in Fig.7. A heatmap is generated by drawing the peak value of the FPRP vector at each reference point in the two-dimensional space. The strongest RSSI is given the red color and the weaker the signal becomes the more the colors move to the blue range. It can be visually seen in Fig.7 that there more parts of the facility covered by the wireless access points signal in the fourth configuration in contrast to the first configuration. The heatmap is just a rough visualization tool for the configuration results. It is not an accurate tool for the evaluation of a configuration since only the peak value of the FPRP is drawn and in order for the positioning system to work, a reference point has to be covered by more than three wireless access points as has been explained in Section 2.2.

Figure 6. Localization Error for Different Configurations

2510



Figure 7: Heatmap of the First and the Fourth Configuration

We have used our environment to compare between the performance of the selection function J from (Ismail et al. 2016) and the selection function f from (Zhao, Yongxiang, Huaibei Zhou 2008). Authors in (Ismail et al. 2016) claim that their selection function J performs better than the selection function f without numerical evidence. However, after using these functions for eight different configurations as shown in Fig.8, results shown a 95% correlation between the selection function J results and the localization error and only 72% correlation between the selection function f and the localization error. However, the correlation results have been performed after removing the anomaly result in the selection function J of the second configuration marked by a red bar. This anomaly resulted because we have not assigned a penalty value P while running the selection function J. Since the selection function J outperforms the selection function f, we use it as a fitness function in the Genetic Algorithm. The Genetic Algorithm runs and generates a set of optimum configurations for different number of wireless access points as shown in Fig.9. It is noticed in Fig.9 that more than 35 access points result in no significant reduction in the localization error and that the error reaches 1.14 meter, which is very close to the ideal one meter lower bound mentioned in Section 3. A detailed view of the optimum configuration is given in the following Table 2 with the indexes of the APs selected by the Genetic Algorithm and the heatmap of the selected configuration. It can be seen from the heatmap how the whole area of the manufacturing facility is covered.

Figure 8. Comparison between Selection Function J and f with Respect to the Localization Error

2511



Figure 9. Six Optimum Configurations for Six Different Number of APs

Table 2. Detailed View of the Optimum Configuration

Number of APs

Lister APs as Output from GA

Best Average Localization Error

Number of dead zones

Signal Coverage

35

7, 48, 69, 26, 65, 61, 74, 10, 28, 49, 53, 37, 79, 67, 55, 13, 38, 44, 36, 9, 54, 87, 19, 64, 79, 62, 14, 46, 81, 86, 42, 68, 70, 23, 72

1.14 0

6. Conclusions and Future Works In this paper, we have discussed the benefits of automating the material handling process and we have proposed a solution using mobile robots system that navigates using existing wireless network infrastructure as the most cost efficient solution. The existing wireless network infrastructure, however, cannot be used with its current configuration since its objective has changed. The wireless network has to provide the robot with accurate information about its location in addition to providing users with network coverage. We have conducted experiments on a large layout of 17, 500m2 similar to a manufacturing facility floorplan using the fingerprinting technique and by changing the wireless access points locations. The localization error has been changed by 46% proving that the localization error can greatly be reduced by only changing the wireless access points locations. Following the above line of thought, an optimization model using Genetic Algorithm has been developed using the selection function J (Ismail et al. 2016) as a fitness since it has not been used a fitness function in its original paper. The Genetic Algorithm managed to find a configuration that guides the robot with a localization error of 1.14 meters only utilizing 35 wireless access points only. It is required as future work to develop a fitness function that can further reduce the localization error. The effect of the fitness function has been investigated by comparing two different selection functions from literature and it has been found that the selection function J correlates to the localization error more than the other function. The optimum configuration is being chosen according to the fitness function and it is sought to have a configuration resulting in minimum localization errors at runtime and therefore the more correlated the fitness function to the localization error the better. Moreover, as future work it is required to study the effect of the variations in the wireless signal on the localization error and enhance the robot offline database by adding the probabilistic parameters to the fingerprints in order to reach localization errors less than one meter.

2512



References Alsmady, A. and Awad, F., Optimal Wi-Fi Access Point Placement for RSSI-Based Indoor Localization Using Genetic

Algorithm. In International Conference on Information and Communication Systems (ICICS), pages 287–291, Irbid, Jordan, (2017).

Baala, O., Zheng, Y., and Caminada, A., The impact of AP placement in WLAN-based indoor positioning system. In 2009 Eighth International Conference on Networks, (2009).

Baggio, A. and Langendoen, K., Monte Carlo localization for mobile wireless sensor networks. Ad Hoc Networks, Available online at www.sciencedirect.com, 6(5):718–733, (2008).

Battiti, R., Brunato, M., and Delai, A., Optimal Wireless Access Point Placement for Location-Dependent Services, Technical report DIT-03-052, Universita` di Trento, October (2003).

Betke, M. and Gurvits, L., Mobile robot localization using landmarks. Robotics and Automation, IEEE International Conf. on Robotics and Automation., pages 135–142, . (1997).

Biswas, J. and Veloso, M., WiFi Localization and Navigation for Autonomous Indoor Mobile Robots. IEEE International Conference on Robotics and Automation, pages 4379–4384, (2010).

Chen, L., Li, B., Zhao, K., Rizos, C., and Zheng, Z., An improved algorithm to generate a Wi-Fi fingerprint database for indoor positioning. Sensors (Basel, Switzerland), 13(8):11085–11096, (2013).

Davich, T., Material Handling Solutions : A look into Automated Robotics, Department of Industrial and Systems Engineering University of Wisconsin-Madison January 9th, (2010).

Dellaert, F., Fox, D., Burgard, W., and Thrun, S., Monte Carlo localization for mobile robots. Proceedings 1999 IEEE International Conference on Robotics and Automation (Cat. No.99CH36288C), 2(May):1322– 1328, (1999).

Farkas, K., Huszak, A .́, and G ́o, G., Optimization of Wi-Fi access point placement for indoor localization. Scientific Publication www.sci-pub.com, ISSN: 1339-147X, Volume 1, Issue 1, July 2013:28–33, (2013).

Gamberi, M., Manzini, R., and Regattieri, A., An new approach for the automatic analysis and control of material handling systems: Integrated layout flow analysis (ILFA). International Journal of Advanced Manufacturing Technology, 41(1-2):156–167, (2009).

Ismail, A. H., Tasaki, R., Kitagawa, H., and Terashima, K., Optimum placement of wireless access point for mobile robot positioning in an indoor environment. Journal of Robotics and Mechatronics, 28(2):162–172, (2016).

Jain, V. K., Tapaswi, S., and Shukla, A., Distributed location estimation system using wlan received signal strength fingerprints. IEEE Wireless Communication and Networking Conference, WCNC, pages 3102– 3106, (2012).

Jing, L. and Yang, P., A Localization Algorithm for Mobile Robots in RFID System. 2007 International Conference on Wireless Communications, Networking and Mobile Computing, pages 2109–2112, (2007).

Kaemarungsi, K. and Krishnamurthy, P., Properties of indoor received signal strength for WLAN location fingerprinting. Proceedings of MOBIQUITOUS 2004 - 1st Annual International Conference on Mobile and Ubiquitous Systems: Networking and Services, pages 14–23, (2004).

J. Kim and W. Chung, Localization of a Mobile Robot Using a Laser Range Finder in a Glass-Walled Environment. IEEE Transactions on Industrial Electronics., vol. 63, no. 6, pp. 3616–3627, 2016.

Ladd, A. M., Bekris, K. E., Rudys, A., Kavraki, L. E., and Wallace, D. S., Robotics-Based Location Sensing Using Wireless Ethernet. Proceedings of the Annual International Conference on Mobile Computing and Networking, pages 227–238, (2002).

Le Dortz, N., Gain, F., and Zetterberg, P., WiFi fingerprint indoor positioning system using probability distribution comparison. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 2301–2304, (2012).

Le Dortz, N. and Gain, F. and Zetterberg, P., WiFi Fingerprint Indoor Positioning System Using Probability Distribution Comparison. Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on, pages 2301–2304, (2012).

Lee, S., Tewolde, G. S., Lim, J., and Kwon, J., Vision Based Localization for Multiple Mobile Robots Using Low-cost Vision Sensor. IEEE International Conference on Intelligent Robots and Systems Conf. on Electro/Information Technology, pages 280–285, (2015).

Li, B., Salter, J., Dempster, A. G., and Rizos, C., Indoor positioning techniques based on wireless {LAN}. Proc. AusWireless 2006, (2006).

Li, R., Du, Z., Zhao, Y., and Liu, S., Design and Implementation of Mobile Robot Ultrasonic Localization System. 28th Chinese Control and Decision Conference (CCDC), pages 5347–5352, (2016).

2513

http://www.sciencedirect.com/

http://www.sci-pub.com/



Liao, L., Chen, W., Zhang, C., Zhang, L., and Jia, W., Wireless access point deployment for both coverage and localization. 2010 IEEE Fifth International Conference on Networking, Architecture, and Storage, pages 348–356, (2010).

Lott, M. and Forkel, I., A multi-wall-and-floor model for indoor radio propagation. In IEEE VTS 53rd Vehicular Technology Conference, pages 464–468, (2001).

M. Youssef, A. Agrawala, A. U. Shankar, and S. H. Noh, “A Probabilistic Clustering-Based Indoor Location Determination System”. Tech. Report, University of Maryland at College Park, CS-TR 4350, March (2002).

Ma, R., Guo, Q., Hu, C., and Xue, J., An improved WiFi indoor positioning algorithm by weighted fusion. sensors ISSN 1424-8220, 15:21824–21843, (2015). Milioris, D., Tzagkarakis, G., Papakonstantinou, A., Papadopouli, M., and Tsakalides, P., Low- dimensional signal-

strength fingerprint-based positioning in wireless LANs. Ad Hoc Networks, 12:100–114, (2014). Mirowski, P., Palaniappan, R., and Ho, T. K. Depth camera slam on a low-cost WiFi mapping robot. In 2012 IEEE

International Conference on Technologies for Practical Robot Applications (TePRA), pages 1–6, (2012). Navarro, E., Peuker, B., Quan, M., Clark, A. C., and Jipson, J., Wi-Fi Localization Using RSSI Fingerprinting. Test,

pages 1–6, (2008). Nerguizian, C. and Outemzabet, S., Accuracy Enhancement of an Indoor Artificial Neural Network- based

Fingerprinting Location System Using Kalman Filtering. IEEE PIMRC 2008 Services, Applications & Business Track, pages 1–5, (2008).

Pandya, D., Jain, R., and Lupu, E., Indoor location estimation using multiple wireless technologies. In 14th IEEE Proceedings on Personal, Indoor and Mobile Radio Communications, PIMRC 2003, volume 3, pages 2208 – 2212, (2003).

Recommendation ITU-R P.1238-1, Propagation data and prediction methods for the planning of indoor radio communication systems and radio local area networks in the frequency range 300 MHz to 100 GHz. Technical report, (2015).

Rencken, W. D., Concurrent Localization and Map Building for Mobile Robots Using Ultrasonic Sensors. Proc. ̃of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS’93), 00(C):2192– 2197, (1993).

Skobeleva, A., Ugrinovskii, V., and Petersen, I., Extended Kalman Filter for indoor and outdoor localization of a wheeled mobile robot. Australian Control Conference November 3-4 , pages 212–216, (2016).

Sobreira, H., Moreira, A. P., and Costa, P. G., Robust Mobile Robot Localization based on Security Laser Scanner. 2015 IEEE International Conference on Autonomous Robot Systems and Competitions , pages 162–167, (2015).

Sun, Y., Liu, M., and Max, M., WiFi signal strength-based robot indoor localization. 2014 IEEE International Conference on Information and Automation, ICIA 2014, (July):250–256, (2014).

Tompkins, J. A., White, J. A., Bozer, Y. A., and Tanchoco, J. M. A., Facilities planning (third edition). Wiley, New York, 3 edition, (2003).

Triggs, B. and Cameron, S., Sonar Localisation for Mobile Robots: A Model-Based Approach. Proceedings of the International Workshop of Emergent Technologies and Factory Automation, pages 387–392, (1992).

Wang, Y., Jia, X., and Lee, H., An indoors wireless positioning system based on wireless local area network infrastructure. In The 6th International Symposium on Satellite Navigation Technology Including Mobile Positioning & Location Services, (2003).

Xiang, Z., Song, S., Chen, J., Wang, H., Huang, J., and Gao, X., A Wireless LAN-based Indoor Positioning Technology. IBM Journal of Research and Development, 48(5/6):617–626, (2004).

Youssef, M., HORUS: A WLAN-based indoor location determination system. Department of Computer Science University of Maryland, (2004).

Zhao, Yongxiang, Huaibei Zhou, M. L., Indoor Access Points Location Optimization using Differential Evolution. International Conference on Computer Science and Software Engineering, pages 382–385, (2008).

Biographies Ahmed F. Abdel Moneim received his PhD in Engineering from Odessa Institute of Marine Engineers in 1970. He was the chairman of the Alexandria Shipyard from 1981 to 1986. He was the chair of the two major departments at the Arab Academy for Science, Technology and Maritime Transport (AAST); the Basic and Applied Sciences Department from 1992-1994, and the Industrial and Management Engineering Department from 1996-1999. During 1994-1996, he was the chairman of the Alexandria Ship Repairing Company. Currently he is a professor at the Industrial and Management Engineering Department and has published many research papers in the field of Reliability Engineering, Project Management and Operations Research. His e-mail address is [email protected].

2514

mailto:[email protected]



Hala A. Farouk received her PhD in Engineering from Cairo University, Egypt in 2011. She has been teaching and conducting research at the Computer Engineering Department in the Arab Academy for Science, Technology and Maritime Transport (AASTMT) from 2001 to 2012. Currently, she is an assistant professor at the Industrial and Management Engineering Department in AASTMT. Her research interests center on hardware design and its adoption in industry, security of hardware systems and the adoption of evolutionary optimization in industry. Her e-mail address is [email protected]. Tarek Ibrahim received his B. Sc. in Engineering from Alexandria University, Electronics and Communication Department, Egypt. He is working in the field of quality and production of Industrial Manufacturing Facilities since 1997. He is studying now for the Master of Science degree in the Industrial and Management Engineering Department at the Arab Academy for Science, Technology and Maritime Transport (AASTMT). His research interests in the field of optimization and simulation of the Manufacturing System. His e-mail address is [email protected].

2515



Optimal Wireless Access Points Placement in a Manufacturing...

Documents

Transcript of Optimal Wireless Access Points Placement in a Manufacturing...