Optimal Policy Determination for Autonomous Underwater Robots Using MDP and POMDP Frameworks