Studying the Effect of Facility Size on the Selection of Automated Guided Vehicle Flow Configurations

Studying the Effect of Facility Size on the Selection of Automated Guided Vehicle Flow Configurations Tarek Al-Hawari, Ena'am S. Al-Zoubi, and Hussam Alshraideh Industrial Engineering Dept., Jordan University of Science and Technology, Irbid, Jordan Email: tarek321@just.edu.jo Abstract—In this paper, simulation is used to compare the performance of three Automated Guided Vehicle (AGV) flow configurations: conventional, tandem and loop in three differently sized facilities. The objective is to study the effect of these configurations in various facilities on minimizing the ratio of total unloaded travel time of all AGVs in the system. It is found that this ratio is highly affected by AGV configurations as well as the size of the facility and number of AGVs used. Index Terms—simulation, AGV, flow configuration, facility layout I. INTRODUCTION A flexible manufacturing system (FMS) is a system composed of machines interconnected with an automated material handling system (MHS). Automated Guided Vehicles (AGV) are a very popular choice as an MHS in FMSs because of their versatility and speed [1]. AGVs were first suggested in 1955 [2] which are transportation vehicles with electronic devices that move around a network along a guide path using special guidance methods and on-board navigation systems [3]. AGVs can be installed in both indoor and outdoor environments, in distribution, shipping, manufacturing and storage areas [4]. The main drawback of using AGVs is their high cost [5]. Flow patterns have a significant effect on travel time, operating expenses and installation costs of the system. Moreover, these patterns have a key role in determining the complexity of the facility's MHS [6]. The most common types of patterns found in literature and industry are three; Traditional, Tandem, and Loop which will be described in the next section. These types of problems have a huge importance in industry, since 20% to 50% of the total operating expenses are spent on material handling and an appropriate layout design can reduce the overall cost by at least 10-30% [7]. According to De Guzman [8] designing the flow path layout is the most important variable in organizing AGV systems. Various solution procedures to the routing problem and flow path comparisons have been reported in literature. These procedures are classified into three principal Manuscript received February 25, 2015, revised May 16, 2015. categories: mathematical procedures, metaheuristic methods, and simulation [9]. Several classes of mathematical (exact) algorithms based on integer programming [10], [11], dynamic programming [12], and graph theory [13] have been used in solving this problem. For example, Fazlollahtabar et al. [14] tried to optimize the material flow in a flexible job-shop automated manufacturing system using mathematical programming. The objective was to optimize the material flow with respect to machine specifications and demand fluctuations. Researchers developed several metaheuristics with different search algorithms such as: Simulated annealing [15], tabu search and genetic algorithms (GA) [16] for the routing problem. They show their superiority over exact methods by being able to solve large complicated combinatorial optimization problems in an efficient way and less time. The final category in solving flow path problems which is the subject of our study is simulation. Simulation techniques have been implemented widely in designing AGV configurations in FMS [17]-[25]. Computer simulation is assumed the most convenient and flexible way to evaluate AGV systems [17]. Modeling AGV systems using simulation is extremely useful since it gives designers the opportunity to choose from and examine many design alternatives, based on set parameters and performance measures. It is capable of modeling complex systems that cannot be modeled using exact or metaheuristic procedures [3], [17]. II. AGV FLOW CONFIGURATIONS An introduction to each type of AGV flow configurations is presented next. A. Traditional/Conventional AGV Configuration Conventional or traditional configuration is the basic layout for AGVs flow paths, in which the vehicle is free to go to any node. All AGVs share the same guide path [26] and the network is reachable from any node. Fig. 1 shows an example of this configuration and points out the directions of travel of AGVs on each path. In this configuration the directions of travel are selected based on minimizing the total travel of loaded vehicles algorithm 132 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering doi: 10.12720/joace.4.2.132-139 formulated by Gaskins and Tanchoco [27], in which only unidirectional paths are allowed. The advantages of traditional configurations according to Ross et al. [18] are:  Fewer pickup/drop-off (P/D) stations.  More flexibility and less sensitivity to AGV breakdowns compared to other configurations.  Fewer AGVs (economic benefits). B. Tandem AGV Configuration Tandem configuration was first introduced by Bozer and Srinivasan [29], [30]; this configuration breaks the system into non-overlapping loops where each loop has its own AGV, which decreases the delays due to blocking or congestion encountered in Traditional configuration [7]. Buffer stations are introduced to exchange the materials between the loops, which increase the complexity of the overall system and introduce the loop workloads balancing problem. Fig. 2 shows an example of the Tandem configuration including the buffer (P/D) areas between adjacent loops for the same facility layout that illustrated the Traditional configuration. Figure 1. Traditional/Conventioal AGV configuration. [28] Figure 2. Tandem AGV configuration. [28] The advantages of the Tandem configuration are listed below [29]:  Less complicated control system for each loop (traffic management).  Reduction in installation costs and effort because of the matching control system in each loop.  Facilitates future expansions.  Effective usage of bi-directional paths for AGVs, as compared with traditional configuration.  The analyst can easily find the optimum location of the stations.  Supports effective use of group technology. The tandem configuration limitations are as follows [29]:  The product will probably be handled by two or more vehicles before reaching its destination.  Need to balance workloads between loops to avoid the bottleneck loop problem.  Requires more space and P/D buffer areas than traditional configuration.  Less response to vehicle breakdowns.  It could cause a routing problem for loads that pass through several loops. C. Single-Loop AGV Configuration The single-loop configuration as shown in Fig. 3 is one loop that passes through all stations. It can be noticed that this configuration is a simpler form of the tandem type. The advantages of the single-loop configuration over tandem configuration are [20]:  Decreases the routing and congestion problem caused by load transfers.  Reduces problems due to AGV breakdowns by directing the loads to the next vehicle until the problem is resolved. The limitation of the single-loop configuration compared with tandem configuration is the congestion and blocking problem because of the allowance of more than one AGV on the same loop. Figure 3. Single-Loop AGV configuration III. LITERATURE REVIEW Many research studies have been conducted to compare between AGV configurations using simulation; Ross et al. [20] compared traditional, tandem and tandem/loop configurations. Results indicated the viability of the tandem configuration across all performance measures considered. Choi et al. [25] introduced simulation models developed in SIMAN to compare traditional and tandem configurations based on several performance metrics. Results indicated that the traditional configuration resulted in better flow time performance while tandem configuration has better throughput performance. Bischak and Stevens [31] conducted a comparative study between conventional and tandem AGV track systems using simulation methods. According to them the tandem system has a higher expected travel time per load thus a greater average time in the overall system because of the multiple loops, while having a lower average time in the system than conventional systems when the loads are delivered within a single loop. In this study a comparative study is performed between the three AGV configurations for three different facilities to take into account the size of the facility and its effect on the performance measures of the system. 133 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering IV. PROBLEM FORMULATION AND ASSUMPTIONS In this paper, a comparative study is conducted between various AGV material flow configurations, and their effects on minimizing the total loaded/unloaded travel time. The throughput of the system is fixed to 10000 parts. Discrete event simulation will be used to evaluate the performance of three facility layouts found in the literature; an 8-station [18], 16-station [30], and 22-station [18] facility layouts, which will be shown later in this section. For each facility; a separate model will be built for each configuration (Tandem, Traditional and Loop). The best choice will be based on the minimum Ratio of Unloaded Travel Time (RUTT) of AGVs as compared to the total travel time which includes both loaded plus unloaded times. A two-way ANOVA will be used to determine the statistical significance of simulation results, the two experimental factors are: 1. The AGV flow configuration: Traditional, Tandem and Loop (3-levels). 2. Number of transporters per loop: One, Two, Three and Five Transporters (4-levels). A total of 12 simulation models will be built for each facility size. For a given facility layout, the appropriate flow pattern will be chosen as the one with the minimum RUTT value. A. Assumptions  Each workstation handles one part at a time;  Vehicles and resources are continuously operational without breakdowns;  Set-up time is included in the operation time;  Vehicles carry single loads.  Vehicles are dispatched based on the shortest travelling distance rule.  Loading and unloading time for the transporters are neglected.  The transportation time between the loops in tandem configuration in the three facilities are assumed to be zero.  The AGVs are assumed to travel at a constant speed of 50 m/hr in all the three systems.  5 types of parts are considered in each facility.  The inter-arrival time of parts follows an exponential behavior.  The processing times follow a fourth-order Erlang distribution, with a mean of 10 min. The Erlang distribution was selected over the exponential distribution since it is more general. [32]  200 minutes were used as a warm-up period during the simulation for the three facilities. This period ensures that all systems reach steady state.  The parts, the inter-arrival and processing time were unified for the same facility size, and were chosen so that each station is visited at least once.  The throughput target will be 10,000 parts, which is the termination condition.  Each model was run for 10 replications. V. SIMULATION MODELS The three facilities with the three configurations were modeled using Arena a popular simulator [33]. Before proceeding to the core of this study, some key information must be outlined. Guided-path AGVs will be used in this study; which are transporters that move along a fixed path and can account for interference with other transporters along the path. Arena represents paths as arcs between intersections, while P/D stations and aisle cross points are represented as intersections. The flow direction of the AGV is indicated by directed arcs between intersections. This representation is a network-based system that is very handy in formulating the guided-path AGVs. [34] A. 8-Station Facility Layout Fig. 4 illustrates the locations of each station in the 8-station facility and the distances between them. All distances are in meters. The figure also shows the order release station where five types of parts (A, B, C, D and E) arrive and the exit system station where the finished products leave. (a) (b) (c) (d) Figure 4. The AGV flow configuration for 8-station facility layout. a) The distances between stations in 8-station facility layout. b) The traditional AGV configuration. c) The Tandem AGV configuration. d) single-loop AGV configuration. [18] Parts b, c and d of Fig. 4 represent the AGV configurations for this facility layout, and Table I shows parts information. TABLE I. JOB TYPE AND ALL RELEVANT DETAILS FOR 8-STATION FACILITY Job Type Routing (sequences) Inter-arrival time (hrs) Processing time(min) Part A 4, 2, 6, 7, 8, exit system Expo(8) ERLA(10,4) Part B 1, 4, 2, 7, 8, exit system Expo(10) ERLA(10,4) Part C 4, 6, 7, 8, exit system Expo(9) ERLA(10,4) Part D 1, 4, 2, 6, 7, 8, exit system Expo(12) ERLA(10,4) Part E 1, 6, 8, exit system Expo(7) ERLA(10,4) B. 16-Station Facility Layout Figs 1, 2 and 3 illustrated the 16-station facility layout, and showed the order release station where five types of parts (part A, B, C, D and E) arrive, and the exit system 134 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering station where the finished parts leave. Table II shows the routing, arrival and processing information of the five parts. C. 22-Station Facility Layout Figs. 5, 6 and 7 represent the three AGV configurations for this facility (traditional, tandem and loop configurations, respectively). TABLE II. JOB TYPE AND ALL RELEVANT DETAILS FOR 16-STATION FACILITY. Job Type Routing (sequences) Inter-arrival time (hrs) Processing time(min) Part A 1, 5, 8, 2, 3, 9, 13, Exit system Expo(8) ERLA(10,4) Part B 1, 4, 14, 15, 7, 10, 12, Exit system Expo(10) ERLA(10,4) Part C 6, 2, 6, 13, 12, Exit system Expo(9) ERLA(10,4) Part D 3, 8, 5, 4, 2, 1, 11, Exit system Expo(12) ERLA(10,4) Part E 7, 11, 14, 1, 10, Exit system Expo(7) ERLA(10,4) Figure 5. Traditional AGV configuration for 22-station layout. [18] Figure 6. Tandem AGV configuration for 22-station layout. The parts sequences, arrival and processing time for this facility are shown in Table III. Figure 7. Single-loop AGV configuration for 22-station layout. TABLE III. JOB TYPE AND ALL RELEVANT DETAILS FOR 22-STATION FACILITY. Job Type Routing (sequences) Inter-arrival time (hrs) Processing time(min) Part A 2, 3, 6, 8, 11, 16, 20, exit system Expo(8) ERLA(10,4) Part B 2, 7, 4, 5, 6, 10, 15, 16, 19, 20, 22, exit system Expo(10) ERLA(10,4) Part C 2, 3, 5, 6, 9, 16, 14, 22, exit system Expo(9) ERLA(10,4) Part D 7, 4, 5, 12, 16, 14, 19, 18, exit system Expo(12) ERLA(10,4) Part E 3, 7, 4, 8, 16, 13, 17, 20, exit system Expo(7) ERLA(10,4) D. Performance Measures in Simulation Models Total work in process inventory of parts was used to determine the warm up period and steady state of each model. The key performance measure that used as a basis for comparison was the ratio of unloaded travel time (RUTT). This represents the ratio of unloaded travel time to the total travel time, which is computed using equation (1): Ratio of unloaded travel time = unloaded travel time (unloaded travel time + loaded travel time) (1) This equation takes into account the cumulative unloaded travel times of all AGVs and parts in the model with respect to the total cumulative travel time whether loaded or unloaded of all AGVs and parts. VI. SIMULATION MODELS RESULTS AND DISCUSSION The objective of this study is to compare RUTT values from the 10 replications of the three AGV configurations in each of the three facilities while maintaining an output of 10000 parts as a target throughput from each system. A. Eight-Station Facility Layout Results Table IV shows the RUTT values and their standard deviations for the 8-station facility. Note that each RUTT value represents the average of 10 replications while varying the number of transporters. 135 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering TABLE IV. RUTT VALUES AND THE STAND. DEV. FOR 8-STATION FACILITY. No. of Transporters AGV Configuration Average RUTT Standard Deviation 1 Transporter Traditional 0.39787 0.000835 Tandem 0.39058 0.000646 Loop 0.40009 0.000861 2 Transporters Traditional 0.24496 0.006137173 Tandem 0.18027 0.003215604 Loop 0.28702 0.004922014 3 Transporters Traditional 0.26631 0.005348 Tandem 0.22927 0.004718 Loop 0.30006 0.004719 5 Transporters Traditional 0.29689 0.002326 Tandem 0.22752 0.003261 Loop 0.3071 0.003647 Fig. 8 is the visual representation of the RUTT values for the three AGV configurations for 8-station facility layout, while varying the number of transporters. As can be seen from the four graphs in Fig. 8, the minimum RUTT value is for the tandem configuration, then comes the traditional configuration and finally the loop configuration has the maximum RUTT values. Note that increasing the number of transporters slightly decreases the values of RUTT for the three configurations. B. Sixteen-Station Facility Layout Results Table V and shows the RUTT values and their standard deviations for the 16-station facility layout while varying the number of transporters. Note that each RUTT value represents the average of 10 replications. According to the graphs in Fig. 9, tandem configuration has the minimum RUTT values. Loop configuration comes next and finally, traditional configuration has the maximum RUTT values. Also, when using single transporter, the three configurations perform the same. Increasing the number of transporters decreases the RUTT values for tandem configuration, while in the other configurations RUTT values almost stay the same. TraditionalTandemLoop 0.40 0.35 0.30 0.25 0.20 TraditionalTandemLoop 0.40 0.35 0.30 0.25 0.20 1 AGV R a t io 2 3 5 Boxplot of Ratio Panel variable: no. of transporters Figure 8. The box plot of RUTT values for 8-station facility layout while varying the number of AGV transporters. TABLE V. RUTT VALUES AND THE STAND. DEV. FOR 16-STATION FACILITY. No. of Transporters AGV Configuration Average RUTT Standard Deviation 1 Transporter Traditional 0.43953 0.00099672 Tandem 0.4443 0.00073786 Loop 0.4337 0.00103602 2 Transporters Traditional 0.43923 0.0092832 Tandem 0.26183 0.00386955 Loop 0.36255 0.00437829 3 Transporters Traditional 0.43205 0.00315885 Tandem 0.27358 0.00180911 Loop 0.3409 0.00358949 5 Transporters Traditional 0.41948 0.00304478 Tandem 0.29051 0.00210895 Loop 0.3319 0.00332365 TraditionalTandemLoop 0.45 0.40 0.35 0.30 0.25 TraditionalTandemLoop 0.45 0.40 0.35 0.30 0.25 1 AGV configuration R a t io 2 3 5 Boxplot of Ratio Panel variable: no. of transporters Figure 9. The box plot of RUTT values for 16-station facility layout while varying the number of AGV transporters. C. Twenty Two-Station Facility Layout Results Table VI shows the average RUTT values and their standard deviations for the 22-station facility. Fig. 10 is the visual representation for RUTT values of this facility. Each part of the figure represents the RUTT values with a different number of transporters. In 22-station facility, tandem configuration performs the best with the minimum RUTT values among the three configurations. Next, loop configuration comes next and finally, traditional configuration has the maximum RUTT values. In this case, increasing the number of transporters decreases the RUTT values by a small amount. On the other hand, increasing the number of transporters doesn't affect the RUTT values for traditional configuration. D. Simulation Results Discussion General observations can be concluded from the three AGV flow configurations for the three facilities:  Tandem configuration shows the best performance (minimum RUTT) in all three facilites. Loop configuration comes next and finally, traditional configuration has the maximum RUTT values. The reason for the good performance of the tandem configuration in all facility sizes is due to its flexibility and ease of control. But the trade-off in 136 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering this configuration is the need for additional buffer areas and AGVs for each loop.  When it comes to traditional configuration; the best performance is at relatively small facilities, and this performance decreases in larger facility sizes.  The performance of loop configuration is intermediate.  Increasing the number of AGVs decreases the RUTT values in an obvious manner. TABLE VI. RUTT VALUES AND THE STAND. DEV. FOR 22-STATION FACILITY. No. of Transporters AGV Configuration Average RUTT Standard Deviation 1 Transporter Traditional 0.59062 0.000590292 Tandem 0.46375 0.000437798 Loop 0.5395 0.001121507 2 Transporters Traditional 0.61457 0.004888774 Tandem 0.23977 0.001984411 Loop 0.46849 0.002029477 3 Transporters Traditional 0.59397 0.002806758 Tandem 0.22814 0.002585515 Loop 0.44391 0.002595059 5 Transporters Traditional 0.56681 0.003654662 Tandem 0.21436 0.001705025 Loop 0.43087 0.001723079 TraditionalTandemLoop 0.6 0.5 0.4 0.3 0.2 TraditionalTandemLoop 0.6 0.5 0.4 0.3 0.2 1 AGV configuration r a t io 2 3 5 Boxplot of ratio Panel variable: no. of transporters Figure 10. The box plot of RUTT values for 22-station facility layout while varying the number of AGV transporters. VII. ANALYSIS OF VARIANCE (ANOVA) RESULTS Two-way ANOVA was conducted. Analysis was carried out for each facility size in MINITAB version 16.1.1. A confidence level of 95 % (i.e. α = 0.05) was used throughout the statistical analysis. The experimental factors taken were: 1. The AGV flow configuration: Traditional, Tandem and Loop (3-levels). 2. Number of transporters: four levels were considered. A single transporter operates for each loop, two, three and five transporters were used. The response is RUTT values. For the evaluation of the significance of the experimental results for the 12 models of each facility, the two-way ANOVA analysis results for each facility are shown in the next sub-sections. A. Eight-Station Facility ANOVA Table VII shows the ANOVA results for the 8-station facility. TABLE VII. THE ANOVA RESULTS FOR 8-STATION FACILITY. Source DF SS MS F P AGV configuration 2 0.091364 0.045682 3909.61 0 no. of Transporters/ Loop 3 0.446048 0.148683 12724.7 7 0 AGV configuration × no. of Transporters 6 0.028261 0.00471 403.11 0 Error 108 0.001262 0.000012 Total 119 0.566934 Fig. 11 represents the residual plots for RUTT values for the 8-station facility. 0.0100.0050.000-0.005-0.010 99.9 99 90 50 10 1 0.1 Residual P e r c e n t 0.400.350.300.250.20 0.008 0.004 0.000 -0.004 -0.008 Fitted Value R e s id u a l 0.0060.0040.0020.000-0.002-0.004-0.006-0.008 20 15 10 5 0 Residual F r e q u e n c y 1201101009080706050403020101 0.008 0.004 0.000 -0.004 -0.008 Observation Order R e s id u a l Normal Probability Plot Versus Fits Histogram Versus Order Residual Plots for RUTT Values Figure 11. Residual plots for RUTT values for 8-station facility. Table VII shows that all main effects and interactions are significant at α = 0.05 significance level, similar tables were generated for the other facilities showing similar results which means that the 12 models for each facility are highly affected by the AGV flow configuration and number of transporters factors. To check for the normality assumption: Fig. 11 represents the residual plots for the 8-station facility. The figure is divided into four plots, the first is the normal probability plot and it resembles a straight line as it should be. The second plot is a plot of residuals versus the fitted values and it doesn't reveal any obvious patterns, which indicates that the variance throughout the data is homogenous. The third plot is a histogram plot for the residuals and it looks like a normal distribution centered at zero (bell shape). The last plot is for residuals versus observation order; no unusual structure is apparent. Similar figures were generated for the other facilities showing similar results. 137 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering These residual plots are used to investigate the normality, variance and independency assumptions in RUTT values for the three facility sizes, they show that assumptions are not violated and that all data follows a normal distribution. VIII. CONCLUSIONS The study shows the effect of AGV flow configurations on the RUTT values of AGVs. It also highlights the dependence of the results on the number of stations and facility layout used. In summary, in small sized facilities the traditional configuration performs the best with respect to RUTT values. All configurations act almost the same in intermediate-sized facilities and in large sized systems; tandem configuration performs the best among all configurations. REFERENCES [1] T. E. Cheng, "AGV dispatching in a flexible manufacturing system," International Journal of Operations & Production Management, vol. 7, no. 1, pp. 62-73, 1987. Business Source Complete, EBSCOhost, viewed 18 May 2014. [2] T. Muller, Automated Guided Vehicles, IFS (Publications) Ltd, Springer-Verlag, UK-Berlin, 1983. [3] B. E. Farling, C. T. Mosier, and F. Mahmoodi, "Analysis of automated guided vehicle configurations in flexible manufacturing systems," INT. J. PROD. RES, vol. 39, no. 18, pp. 4239-4260, 2001. [4] H. Fazlollahtabar and M. Saidi-Mehrabad, “Methodologies to optimize automated guided vehicle scheduling and routing problems: A review study,” Journal of Intelligent & Robotic Systems, Dec. 2013. [5] A. Kahraman, A. Gosavi, and K. Oty, "Stochastic modeling of an automated guided vehicle system with one vehicle and a closed-loop path," IEEE Transactions on Automation Science & Engineering, vol. 5, no. 3, pp. 504-518, 2008. Business Source Complete, EBSCOhost, viewed 18 May 2014. [6] R. S. Putrus, "Layout design: Key to advance assembly/manufacturing success," in Proc. the 4th International Conference on Automated Guided Vehicle Systems, Chicago, IL, 1986, pp. 141-154. [7] D. Sinriech, "Network design model for discrete material flow systems: A literature review," Int. J. Adv. Manuf. Technol, vol. 10, pp. 277-291, 1995. [8] M. De Guzman, N. Prabhu, and J. Tanchoco, "Complexity of the AGV shortest path and single-loop guide path layout problems," International Journal of Production Research, vol. 35, no. 8, pp. 2083-2092, 1997. Business Source Complete, EBSCOhost, viewed 18 May 2014. [9] S. P. Singh and R. R. K. Sharma, "A review of different approaches to the facility layout problems," Int. J. Adv. Manuf. Technol, vol. 30, pp. 425–433, 2006. [10] A. I. Correa, A. Langevin, and L. M. Rousseau, "Scheduling and routing of automated guided vehicles: A hybrid approach," Computer Operation Research, vol. 34, pp. 1688–1707, 2007. [11] G. Desaulniers, A. Langevin, D. Riopel, and B.Villeneuve, "Dispatching and conflict-free routing of automated guided vehicles: an exact approach," Int. J. Flex. Manuf. Syst, vol. 15, pp. 309–331, 2003. [12] Langevin, D. Lauzon, and D. Riopel, "Dispatching, routing and scheduling of two automated guided vehicles in a flexible manufacturing system," Int. J. Flex. Manuf. Syst., vol. 8, pp. 246– 262, 1996. [13] J. Yoo, E. Sim, C. Cao, and J. Park, "An algorithm for deadlock avoidance in an AGV System," Int. J.Manuf. Technol, vol. 26, pp. 659–668, 2005. [14] H. Fazlollahtabar, B. Rezaie, and H. Kalantari, "Mathematical programming approach to optimize material flow in an AGV-based flexible jobshop manufacturing system with performance analysis," Int. J. Adv. Manuf. Technol, vol. 51, no. 9–12, pp. 1149– 1158, 2010. [15] R. Tavakkoli-Moghaddam, M. B. Aryanezhad, H. Kazemipoor, and A. Salehipour, "Partitioning machines in tandem AGV systems based on “Balanced flow strategy” by simulated annealing," Int. J. Adv. Manuf. Technol, vol. 38, pp. 355–366, 2008. [16] R. Z. Farahani, G. Laporte, E. Miandoabchi, and S. Bina, "Designing efficient methods for the tandem AGV network design problem using tabu search and genetic algorithm," Int. J. Adv. Manuf. Technol., vol. 36, pp. 996–1009, 2008. [17] G. C. Vosniakos and A. G. Mamalis, “Automated guided vehicle system design for FMS applications,” International Journal of Machine Tools and Manufacturing, vol. 30, pp. 85-97, 2007. [18] E. A. Ross, F. Mahmoodi, and C. T. Mosier, “Tandem configuration automated guided vehicle systems: A comparative study,” Decision Sciences, vol. 27, pp. 181-102, 1996. [19] B. Mahadevan and T. Narendran, “Design of an automated guided vehicle-based material handling system for a fliexible manufacturing system,” International Journal of Production Research, vol. 28, pp. 1611-1622, 1990. [20] H. G. Choi, H. J. Kwon, and J. Lee, “Traditional and tandem AGV system layouts: A simulation study,” Simulation, vol. 63, pp. 85-93, 1994. [21] J. J. Bartholdi III and L. K. Platzman, “Decentralized control of automated guided vehicles on a simple loop,” IIE Transactions, vol. 21, pp. 76-81, 1989. [22] C. W. Kim and J. M. Tanchoco, “Operational control of a bi-directional automated guided vehicle system,” International Journal of Production Research, vol. 31, pp. 2123-2138, 1993. [23] J. T. Lin, C. C. K. Chang, and W. C. Liu, “A load-routing problem in a tandem configuration automated guided-vehicle system,” International Journal of Production Research, vol. 32, pp. 411-427, 1994. [24] J. Lee, “Composite dispatching rules for multiple-vehicle AGV systems,” Simulation, vol. 66, pp. 121-130, 1996. [25] M. Ozden, “A simulation study of multiple-load-carrying automated guided vehicles in a flexible manufacturing system,” International Journal of Production Research, vol. 26, pp. 1353-1366, 1988. [26] B. E. Farling, C. T. Mosier, and Mahmoodi, “Analysis of automated guided vehicle configurations in flexible manufacturing systems,” International Journal of Production Research, vol. 39, no. 18, pp. 4239–4260, January 2001. [27] R. Gaskins and J. Tanchoco, “Flow path design for automated guided vehicle systems,” International Journal of Production Research, vol. 25, no. 5, pp. 667, May 1987. [28] S. C. Rim and Y. A. Bozer. “The bidirectional circular layout problem,” Working Paper, Industrial Engineering Program, FU-20, University of Washington, Seattle, WA, 98195, 1991. [29] Y. A. Bozer and M. M. Srinivasan, “Tandem configurations for AGV systems offer simplicity and flexibility,” Industrial Engineering, vol. 21, pp. 23-27, 1989. [30] Y. A. Bozer and M. Srinivasan, “Tandem configurations for automated guided vehicle systems and the analysis of single vehicle loops,” IIE Transactions, vol. 23, pp. 72-82, 1991. [31] D. Bischak and K. Stevens Jr. “An evaluation of the tandem configuration automated guided vehicle system,” Production Planning & Control, vol. 6, no. 5, pp. 438. [32] C. W. Zobel and K. B. Keeling, “Neural network-based simulation meta-models for predicting probability distributions,” Computers and Industrial Engineering, vol. 54, no. 4, pp. 879–888, 2008. [33] W. D. Kelton, R. P. Sadoski, and D. A. Sadoski, Simulation with Arena, New York: McGraw-Hill, 1998. [34] S. E. Kesen and O. F. Baykoc, “Simulation of automated guided vehicle (AGV) systems based on Just-In-Time (JIT) philosophy in a job-shop floor environment,” Simulation Modeling Practice & Theory, vol. 15, no. 15, pp. 272-284, 2007. 138 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering Tarek Al-Hawari holds a PhD degree in industrial Engineering from Lehigh University, PA, USA, 2005. Dr. Al-Hawari has been a member of the faculty of engineering at Jordan University of Science and Technology since 2005. He is now an associate professor of industrial engineering and has previously held the position of chairman of the department for a period of 2 years. His research interests include simulation, supply chains and multi-criteria decision making methods. Ena'am S. Al-Zoubi is a Master's degree student at Jordan University of Science and Technology majoring in Industrial engineering. Hussam Alshraideh is an assistant professor of industrial engineering at Jordan University of Science and Technology in Irbid, Jordan. His research interests include data mining techniques and applied statistics. 139 Journal of Automation and Control Engineering Vol. 4, No. 2, April 2016 ©2016 Journal of Automation and Control Engineering