Discrete hidden Markov model based learning controller for robotic disassembly

Abstract

In this paper, a learning controller is constructed by using a discrete Hidden Markov Model (HMM). The HMM states are defined based on the external force/torque sensor information, and the HMM obversation symbols are defined by the velocity command information. In the learning phase of the HMM controller, the Baum-Welch method is used based on the training data obtained from some successful robotic disassembly experiments using heuristic rules defining the desired behavior. During the control of robotic disassembly, the HMM controller can generate the velocity commands based on the most likely performance criterion. Keywords: Learning Controller, Hidden Markov Model, Vector quantization, Robotics 1 Introduction An HMM is a doubly stochastic process with an underlying stochastic process that is not observable (i.e. hidden), but can be observed through another set of stochastic processes that produce the sequence of observed symbols. More detailed reference on theory and compution of H...