Maze is a complex environment where finding optimal path is always a challenge. Recently, traditional Q-Learning and Dyna-CA appear as an effective tool to solve such problems. But major shortcoming with these reinforced learning techniques is that they are not effective when the dimension count of the possible states and actions are relatively high. In such scenario Rule Interpolation based Q-Learning (FRIQ) may be an effective tool to solve the maze problems. Further, MATLAB ® is a computational platform having effective environment to simulate static as well as dynamic maze environment. This research article theoretical analyze all these issues and based upon it decide a research direction for agent based optimization using reinforcement learning in maze environment.