A crash avoidance system for automobiles is developed based upon a distributed network of artificial neurons that mimic the neural organization of the escape system in the American cockroach. The cockroach escape circuit is shown to be an excellent source of inspiration for the development of a collision avoidance system. The crash avoidance system is implemented in an artificial neural network which is trained off-line, but then is shown to produce real-time performance. A collision avoidance scheme which makes use of a crash alarm strategy is developed for training the neural network. A dynamic model of a four-wheeled vehicle with front wheel steering and realistic performance constraints is used to test the crash avoidance system. Simulation results show that the well-trained neural network causes successful, reflexive crash avoidance behaviors in a dynamic environment without a priori information.
|頁（從 - 到）||2007-2012|
|期刊||Proceedings - IEEE International Conference on Robotics and Automation|
|出版狀態||已發佈 - 1997 一月 1|
|事件||Proceedings of the 1997 IEEE International Conference on Robotics and Automation, ICRA. Part 3 (of 4) - Albuquerque, NM, USA|
持續時間: 1997 四月 20 → 1997 四月 25
ASJC Scopus subject areas