✍ Scribed by H. Herbstreith; L. Gmeiner; P. Preuß
No coin nor oath required. For personal study only.
An investigation of a neurally controlled vehicle in a computer-simulated parcours with dynamically changing obstacles is presented. The purpose was to judge the applicability of commercially available neural net shells in process control and automation. The chosen shell provides capabilities for associative memories by unsupervised learning, and for supervised learning by means of the back-propagation algorithm. This algorithm is furthermore enhanced by the funetionallink approach of Pao ([Pao89)). The vehicle and its environment are displayed graphically. The task for the vehicle is to find its way from a user-defined starting point to an ending point. The neural net is responsible for control of the alternating behaviors of target-orientation and obstacle avoidance. We use ten input neurons, nine representing sensors that deliver information about the distance from non-passable areas in any direction. The tenth sensor is responsible for locating the target in a compass-like way. Three output neurons determine one out of seven possible steering directions. The network was trained off-line, with patterns generated schematically by a program. The results are discussed and further refinements proposed.
📜 SIMILAR VOLUMES
A direct adaptive neural control scheme with single and double integral-plus-state (IPS) actions is proposed. The control scheme contains two recurrent trainable neural network (RTNN) models, which are a plant parameter identifier and state estimator, an IPS feedback /feedforward controller, and one