Optimal movement through a maze by an application of Q learning

Machine learning has become a very important aspect of the development of artificial intelligence (AI) in the world of computing. As a part of the machine learning techniques, reinforcement learning enables an agent to explore and study the unknown environment and through trial and error, the agent learns how to perform desired tasks in an optimum way. In this project, one of the methods to solve reinforcement problems, namely Q-learning, is studied and analysed. A maze problem is set up to count the learning time of an agent to travel from a fixed starting point A to destination B using Q-learning algorithm on Matlab platform. The learning time will be studied and then compared with different sets of maze of different size and complexity. Finally, a thorough discussion on the differences in the learning time and methods to further optimise the Q-learning algorithm will be provided. This document reports the outcome of this project and highlights the problems and challenges encountered and solutions devised.