Acquisition of Learning Processing in a Navigation Task Using a Functional Parts Combination Model

摘要

In past studies of intelligent systems for problem solving, the internal structure of the system that determines the output for problem solving, that is, the computation process, is confined to the process built in by the designer. Thus, the program must be rewritten manually, even for a problem of similar type which can be handled by modifications. To deal with this point, we propose the functional parts combination (FPC) model, which can acquire different computation processes according to the individual problem. With the navigation of a mobile agent as a sample problem, the search power of the FPC model for the computation process is demonstrated by computer simulation for the case in which systems with different structures are applied to the same problem, and for the case in which a system with the same structure is applied to multiple problems.