Larry D. Pyeatt and Adele E. Howe
Our focus is on designing adaptable agents for highly dynamic environments. We have implemented a reinforcement learning architecture as the reactive component of a two layer control system for a simulated race car. We found that separating the layers has expedited gradually improving performance. We ran experiments to test the tuning, decomposition and coordination of the low level behaviors. Our control system was then extended to allow passing of other cars and tested for its ability to avoid collisions. The best design used reinforcement learning with separate networks for each action, coarse coded input and a simple rule based coordination mechanism.