Navigation Algorithms in Asteroids

This paper reports on the design, implementation and evaluation of a controller for the classic arcade game Asteroids. The proposed control consists of four submodules: a fuzzy system based fire control for selecting asteroids to target and calibrating the gun on chosen targets, a potential field based navigation control for steering the ship to safer areas, a look-a-head based emergency control for steering the ship out of chaotic situations, and, finally, a fuzzy system based high-level control for switching between the other three submodules. The combined control and its submodules are evaluated, and trained if applicable, using survival time in randomized and preset asteroid fields. Results show that there were clear exploitable weaknesses in the individual submodules where certain preset asteroid fields are far easier to navigate with one module over the other. The combined control did not prove to be far more versatile than the individual navigational modules, although some potential was shown. This is likely due to a lack of training, or parameters to train, from the inherent complexity of simulating the game. The fuzzy system does however show great promise, and with enough optimization it could be a viable solution for similar navigation problems.