Learning to predict Resistive Forces During Robotic Excavation

with the world as excavation. In order to effectively plan its actions, our robot excavator requires a method that allows it to predict the resistive forces experienced as it scoops soil from the terrain. In this paper we present methods for a robot to predict the resistive forces and to improve its predictions based on experience. We start with a simple analytical model of a flat blade moving through soil and show how this analysis can be extended to account for the phenomena specific to excavation. In addition, we examine how representation of the learning problem and methodology affect prediction performance using several criteria.