Distributed planning and control of intelligent robot's arm motion based on symbolic and neural processing

In this work, we present our view of an intelligent robotic manipulator that can work autonomously in a dynamically changing environment. The increased intelligence is needed to enable the manipulator to react to changing environments without contacting a supervising coordinator. This approach requires increased computational capabilities on the part of the manipulator. We show how the computational requirements of such an autonomous manipulator can be satisfied by a combination of neural network-based processing modules. We use symbolic computation methods to construct these neural networks (thereby eliminating training times), and sensors to provide the signals to recognize changes in the environment.