Posterior Probability Estimation Techniques Embedded in a Bayes Filter for Vibration-Based Terrain Classification

摘要: Vibration signals acquired during robot traversal provide enough information to yield a reliable prediction of the current terrain type. In recent approach, we combined history class estimates into final prediction. We therefore adopted Bayes filter taking posterior probability each account. Posterior estimates, however, were derived from support vector machines only, disregarding capability other classification techniques these estimates. This paper considers classifiers be embedded our scheme, featuring different characteristics. show that best results are obtained using k-nearestneighbor and machine approach which has not been considered for so far. Furthermore, demonstrate also benefit temporal filtering predictions.