An MDP-based approach oriented optimal policy for path planning

In this article we present an approach to improve the execution time of the Markov decision process (MDP) used in robotics for path planning. We've improved it for both value iteration algorithms (value iteration) and Policy Iteration (policy iteration). Unlike the conventional approach which initializes the algorithms with random values and explores all the accessible states at each iteration, our approach modifies the classical algorithms to improve their performance. These changes are reflected mainly in two points: non-random initialization of the algorithms, and a significant reduction of calculations at each iteration. The results are promising; they are presented at the end of the paper.