Multi-Agent Simulation Using Adaptive Dynamic Programing for Evaluating Urban Consolidation Centers

This chapter presents details related to the development of a multi-agent system (MAS) with adaptive dynamic programing (ADP) (MAS-ADP). It investigates the performance of the ADP for evaluating urban consolidation centers (UCC) by comparing it with Q-learning. It was found that ADP performed better in all evaluation criteria (accuracy, stability and adaptability, and profitability) when compared with Q-learning. In addition, ADP is more adaptive to the changing environment and it is more stable in the optimal action selection. The MAS-ADP could be used as a decision support tool in city logistics measures to achieve better outcomes. All simulations were done in MATLAB with the different settings of learning rates and discount factors for ADP and Q-learning models based on the results of a sensitivity analysis that has been done prior to the simulation case study.