On time-memory trade-off for collision detection

摘要: Future collision avoidance systems, which are capable of fully controlling the vehicle, have to make critical decisions in a very short time. To do this, they need check constantly if their own vehicle's occupancy collides with other traffic participants' occupancy. Those checks consume substantial amount time and consequently, systems could fail intervene complex scenarios. We propose new approach reduce computation for significantly. Instead using geometric methods, we store finitely many possible scenarios between two objects table thus become matter lookup queries. ensure that finite number configurations cover all scenarios, use novel abstraction technique guarantees every will be detected. The works arbitrarily participants by applying pairwise (own vehicle object) each participant. Randomly generated show can several times faster than intersection techniques thanks trade-off memory consumption