Multi-Agent Based Chess Move Generator System: Taking into Account Local Environments

Chess has spawned a great deal of successful research in the artificial intelligence (AI) community. The strength of some chess programs can be compared to top-level human players. Nevertheless, most of the AI research has been focused on the refinement of evaluation functions which are used to search the NP-complete solution space. Some optimizations have also been made to the arrangement in which possible moves are examined by the search algorithm. In this article we take a nouvelle approach to the generation of chess moves with the help of multi-agent based systems. We design a multi-agent system which represents the pieces of a chess battlefield and we feed the search algorithm with the most promising move(s) from the point of view of the individual agents. Thus, we take into account the individual decisions (tactics) to define the strategy, as well as cutting out the search space. We plan to make a prototype implementation of the multi-agent based chess move generator system (MAB-CMGS) using the FIPA-OS architecture to test the overall system behaviour and achievable chess power.