RAPID CONCEPT-LEARNING FOR MOBILE ROBOTS

Concept learning in robotics is an extremely challenging problem: sens ory data is often high dimensional, and noisy due to specularities and other irregularities. In this paper, we investigate two general strat egies to speed up learning, based on spatial decomposition of the sens ory representation, and simultaneous learning of multiple classes usin g a shared structure. We study two concept learning scenarios: a hallw ay navigation problem, where the robot has to induce features such as ''opening'' or ''wall''. The second task is recycling, where the robot has to learn to recognize objects, such as a ''trash can''. We use a common underlying function approximator in both studies in the form of a feedforward neural network, with several hundred input units and mu ltiple output units. Despite the high degree of freedom afforded by su ch an approximator, we show the two strategies provide sufficient bias to achieve rapid learning. We provide detailed experimental studies o n an actual mobile robot called PAVLOV to illustrate the effectiveness of this approach.