Evolving the optimal racing line in a high-end racing game

Finding a racing line that allows to achieve a competitive lap-time is a key problem in real-world car racing as well as in the development of non-player characters for a commercial racing game. Unfortunately, solving this problem generally requires a domain expert and a trial-and-error process. In this work, we show how evolutionary computation can be successfully applied to solve this task in a high-end racing game. To this purpose, we introduce a novel encoding for the racing lines based on a set of connected Be #x00B4;zier curves. In addition, we compare two different methods to evaluate the evolved racing lines: a simulation-based fitness and an estimation-based fitness; the former does not require any previous knowledge but is rather expensive; the latter is much less expensive but requires few domain knowledge and is not completely accurate. Finally, we test our approach using The Open Racing Car Simulator (TORCS), a state-of-the-art open source simulator, as a testbed