Multi-agent learning for adaptive scheduling

This dissertation incorporates a summary of the results from the simulation based study of multi-agent learning in Distributed Artificial Intelligence (DAI) framework. Specifically, the author proposes and tests different coordination strategies for different possible manufacturing shop floor configurations. Results from the simulation based learning in DAI are bench marked against the results from independent uncoordinated learning and heuristic First Come First Serve (FCFS) approaches for the same set of conditions. The proposed coordination strategies are tested for two different managerial objectives (flow time minimization and tardiness minimization) for their robustness. The results show that distributed learning performs better than both independent agent uncoordinated learning and heuristic FCFS approaches for scheduling jobs in a shop floor environment.