Integration of knowledge-based system and neural network techniques for autonomous learning machines

The authors have developed an automated approach in which a rule-based system supervises the training of a neural network and controls the operation of the system during the learning process. For a preliminary demonstration of these concepts, a simulation in which a two-link manipulator is taught how to make a tennis-like swing has been constructed. The control system first determines how to make a successful swing using rules alone. It then teaches a neural network how to accomplish the task by having the network observe and generalize on rule-based task execution. Following initial training, a rule-based execution monitor evaluates the neural network performance and reengages rule-based swing-maneuver control whenever errors due to changes in the manipulator or its operating environment necessitate retraining of the network. The rule-based system thereby ensures proper task completion while neural network relearning takes place. The simulation shows the interaction between rule-based and network-based system components during various phases of training and supervision.<>