| In tandem Automated Guided Vehicle (AGV) systems, the shop floor is partitioned into a group of non-overlapping zones, each served by a single dedicated AGV. Pickup/drop-off (P/D) points are installed to link these zones as transfer points. In this thesis, a genetic algorithm (GA) is proposed for partitioning the tandem AGV systems. The objective is to minimize the maximum AGV workload in order to balance the workload among all the zones to avoid the occurrence of bottlenecks. The performance of the proposed algorithm is evaluated through comparison with the reported results in the literature. The results show that the performance of the proposed algorithm is superior compared to the algorithms reported in previous studies.;The difficulty of applying the GAs to a practical problem is tuning up their parameters such as population size, crossover rate and mutation rate. The performance of a GA is strongly affected by the chosen values of these parameters. In this thesis, Design of Experiments (DOE) is used to define the best combination of the developed GA parameters' values by analyzing the main effect of each parameter and interaction effects between these parameters and some system characteristics on the obtained solutions' quality and computational time. The considered system characteristics are system size, expected zone loading, and the designated number of zones. The obtained results demonstrated the efficiency of the presented systematic method of tuning the GA's parameters to solve the partitioning problem of tandem AGV systems. A local search algorithm is then proposed and combined with the developed GA to improve its performance. Hence, a new memetic algorithm (MA) is proposed and applied to optimize the partitioning problem of the tandem AGV systems. Then a performance comparison between the developed GA and MA is then carried out on a group of benchmarking problems. The obtained results demonstrated the efficiency of the developed MA in solving the partitioning problem of tandem AGV systems. In terms of solutions quality, the proposed MA outperforms all the previous approaches as well as pure GA. On the other hand, MA seems worse than pure GA in terms of computational time, especially for large size problems. However, the computational time of MA is still within the accepted range. As for the vehicle dispatching problem in AGV systems, a simulation study combined with experimental design is conducted to analyze the effects of a number of empty vehicle dispatching rules which are Shortest Time to Travel First (STTF), First-Encountered-First-Served (FEFS), Largest Queue Size (LQS), and First-Come-First-Served (FCFS). Two configurations of a benchmark problem are simulated with the mentioned dispatching rules on three system performance criteria: average vehicle workload, throughput rate, and average queue length. |
