基于Q-learning算法和ε-greedy策略解决随机生成的方形迷宫问题Matlab代码实现.rar

# Square Random Generated Maze Solving Using Q-learning Algorithm and epsilon greedy policy The aim of this code is solving a randomly generated square maze (dimension ) using a Q-Learning algorithm involving an epsilon greedy policy. The user is capable of defining 3 parameters: - Maze Dimension - Starting point - Ending point Firstly, the code will generate a random maze matrix based on those inputs. Then, so as to apply the Q-Learning algorithm, a reward matrix is built distinguishing between wall-cells (extremely low reward), non-wall-cells, start and end locations. Afterwards, a matrix containing the Q-values of all the possible actions is computed (this is the essence of the Q-Learning algorithm). As it is discussed deeply on the report (Maze_Q_Learning_Report.mlx), this process involves and epsilon greedy policy among other significant factors. Finally, the optimal path is defined.