ADVANCED TELEROBOTICS CONCEPTS USING NEURAL NETWORKS

WE INTEND TO APPLY NEURAL NETWORK TECHNOLOGY TO TELEROBOTICS. THIS WILL TAKE ADVANTAGE OF SOME OF THE ADVANCES THAT ARE BEING DONE FOR AUTONOMOUS ROBOTS. THESE CONCEPTS WILL EXTEND THE CAPABILITIES OF THE HUMAN OPERATOR. THE TELEROBOTIC CONTROLLER WILL BE TAUGHT USING BACKWARD SHAPING TO IMPROVE ITS SPEED OF LEARNING. IT IS PROPOSED TO ADD A CAPABILITY FOR ASSOCIATION BASED SENSOR DATA FUSION. THE CONTROLLER WILL BE ADAPTED TO HAVE MULTIPLE JOINTS CONTROLLED BY HYBRID NEURAL NETWORKS. THE PROJECT WILL SEE IF THE NEURAL NETWORK CAN IMPROVE PERFORMANCE FOR MINIMUM JERK AND MINIMUM TORQUE CRITERIA BASED UPON THE LEARNING. IT IS THOUGHT THAT THE CONCEPTS WILL ALLOW FOR CONTROL OF MULTIPLE JOINTS OR DEGREES OF FREEDOM IN THE ROBOTIC PROBLEM.