A Hybrid Connectionist/LCS for Hidden-state Problems

This data includes the evaluation of TRACA with other learning algorithms two maze navigation tasks. Specifically comparisons are made with other LCS based approaches including XCSMH and AgentP. Each algorithm is evaluated using a maze navigation task that has been identified as among the most difficult due to recurring aliased regions. The comparisons between algorithms include training time, test performance, and the size of the learned rule sets. TRACA is also tested on two variations of a truck driving task where it must learn to navigate four lanes of slower vehicles whilst avoiding collisions.