NEURAL NETWORKS FOR VEHICLE PATH PLANNING AND CONTROL

摘要

This paper presents an algorithm that allows a mobile robot to safely navigate in an unknown environment. The parallel characteristics of the computations involved suggest the use of a distributed processing paradigm. Local optimization is used to favor fast response, while global optimization is pursued by making the navigation algorithm a monotone decreasing function of environment knowledge. The robot configuration space takes into account both the minimum path to the goal and the safest path among obstacles. The network is also robust to environment uncertainties.