We introduce a novel control architecture for Autonomous Mobile Robots called the Reflexive Instructor (RI) with Deliberate Apprentice (DA). The architecture employs simple reinforcement signals provided by the RI component to train the DA. The DA is responsible for providing control signals to the agent’s actuators based on received sensor input. The RI provides a measure of safety in this respect as it is responsible for taking over control of the mobile robot if the DA makes a mistake as well as providing an appropriate feedback signal to the DA. The RIDA interaction is advantageous because it protects the vehicle from its own incompetence and has the potential to accelerate learning in the DA. We illustrate this by simulating a vehicle employing a simple RI coupled to a rapid reinforcement artificial neural network as a DA. The DA learns to use sensors while successfully interacting with its environment.
Proc. of the 3rd IMACS/IEEE International Multiconference on: Circuits, Systems, Communications and Computers (CSCC’99), Athens Greece July 4-8, 1999