Tuning a fuzzy racing car by coevolution

In this paper, we design, build and tune a fuzzy rule-based car controller for FUZZ-IEEE 2007 Car Racing Competition. The membership functions of the car controller are tuned with coevolutionary genetic algorithms. Cooperative and competitive approaches to tuning parameters are compared. In principle, results obtained with a cooperative approach with a BLX cross operator are slightly better than results derived from a competitive method with a 1-point cross operator. In any case, further experiments are needed to support our findings.