A controller for a self-driving racing car in the TORCS platform, implemented as part of the UvA/VU course Computational Intelligence (Fall 2016). The controller was trained with a Multilayer Perceptron using the DL4J Java library. The resulting weights were optimized by means of differential evolution. See report or presentation. Project won the first prize in the annual Computational Intelligence contest associated with the course.
