Decomposition algorithm for performance evaluation of AGV systems

This paper proposes a realistic queueing model of automated guided vehicle (AGV). systems in just-in-time production systems. The model takes into consideration return paths, Erlang distributed service times, and pull-type dispatching rule, assuming finite buffer capacities. Since it has no product-form solution and natural decomposability due to complex nontree fork-cum-join architecture and dynamic dispatching rules, we propose a machine-based decomposition algorithm for the performance evaluation of the model. Each decomposed module consists of the processing machine and its dispatching station. Three flow probabilities, derived from flow conservation analysis, relate the modules, which are updated iteratively until the parameters converge. The numerical results from a real-life AGV system application show that the algorithm is reasonably accurate.