We introduce an application-oriented reinforcement learning searching algorithm designed for problem with fast learning and capturing goal in less amount of time especially in robotics and games. The importance of game playing in machine learning is a...
We introduce an application-oriented reinforcement learning searching algorithm designed for problem with fast learning and capturing goal in less amount of time especially in robotics and games. The importance of game playing in machine learning is an exhaustive application of autonomous agent in real-world problem domain. In our previous published article represent that how autonomous agent learned through self-training and successful trained agent ready for execution [11].In this paper, we design and proposed a new application-oriented searching algorithm especially for game playing in grid world problem. In which first of all agents train all state and able to capture goal successfully. Reinforcement learning is a type of decision making system that takes decision on the basis of reward or penalty signal and learned from environment. Many games, there are no such things that follow fast learning as well as searching and genuine movement for each step. For every state action agent stored previous values in terms of q values in a look-up table. It helps for agent decision making capability during goal hitting or pray captured in the real-world game. In order to access and simulate new searching algorithms in mat lab and evaluated by comparison with different RL techniques [2, 11-12].