The performance of fuzzy rule-based systems are examined in a car racing domain. In the car racing domain, two car agents compete for waypoints. Fuzzy rule-based systems are used to determine which waypoints the car agent should aim at. Training patterns for constructing fuzzy rule-based systems are generated through preliminary experiments against a fixed opponent car agent. We examine the effect of the opponent car agents on the performance of the fuzzy rule-based systems.
展开▼