Intelligent exploration of unknown environments with vision like sensors

In this work we present a methodology for intelligent path planning in an uncertain environment using vision like sensors. We show that the problem of path planning can be posed as the adaptive control of an uncertain Markov decision process. The strategy for path planning then reduces to computing the control policy based on the current estimate of the environment, also known as the "certainty equivalence" principle in the adaptive control literature. We propose a Monte Carlo based estimation scheme, incorporating non local sensors, for estimating the probabilities of the environment process, which significantly accelerates the convergence of the associated path planning algorithms