Two-timescale learning using idiotypic behaviour mediation for a navigating mobile robot

A combined short-term learning (STL) and long-term learning (LTL) approach to solving mobile-robot navigation problems is presented and tested in both the real and virtual domains. The LTL phase consists of rapid simulations that use a genetic algorithm to derive diverse sets of behaviours, encoded as variable sets of attributes, and the STL phase is an idiotypic artificial immune system. Results from the LTL phase show that sets of behaviours develop very rapidly, and significantly greater diversity is obtained when multiple autonomous populations are used, rather than a single one. The architecture is assessed under various scenarios, including removal of the LTL phase and switching off the idiotypic mechanism in the STL phase. The comparisons provide substantial evidence that the best option is the inclusion of both the LTL phase and the idiotypic system. In addition, this paper shows that structurally different environments can be used for the two phases without compromising transferability.