Navigation Under Obstacle Motion Uncertainty Using Markov Decision Processes