Imaging And Controls For Mars Robots With Neural Networks

Abstract

Two aspects of the design of space robots is covered implemented by neural networks and by hybrid approach with artificial intelligence. One is a neurocontroller for a real-time autonomous system. An optical control system developed saves the time for the image processing that analyzes an image sensor through the environment and induces a transformation over the sensor array. A prototype of the neurocontroller is able to learn and control by itself. The second aspect deals with the design of a Servo Control System for a Robot with the capability of "learning in Unanticipated Situations" incorporated in the system. The robot is assumed to be employed to perform useful tasks in an alien evironment. The model developed is shown to provide the robot with the capability to recover from unanticipated situations that can lead to the disruption of its normal operation, and to learn to avoid such situations in the future. These two aspects will be integrated for a design of a very intelligent autonomous space robot.