Autonomous Rover Control with Behavioral Algorithms and Neural Networks

An autonomous control system of a rover for planetary exploration can be subdivided in two distinct parts, the first devoted to planning and scheduling and the second to the plan execution. In this paper the concept of a hybrid intelligent system composed by three layers will be presented, focusing on the plan execution part which is embodied by a Behavioral and a Reflexive layer. The plan generated by the Planning layer is composed by a sequence of logical actions that can not be directly executed by the rover hardware. The Behavioral layer translates these actions in executable actions through a Fuzzy subsystem, and prepares the input for the Reflexive layer. This layer is in charge of the low level control of the rover hardware which is performed by an Artificial Neural Network. In this paper, the whole Execution system will be presented, focusing on the path following problem and reporting the results of the first tests executed on a real prototype.