Q-learning is a kind of model-free reinforcement learning algorithm which is effective in Robot's navigation applications. Unfortunately, Lego Mindstorms EV3 robot's file writing speed is sometimes too slow to implement Q-learning algorithm. In this paper, an approach is proposed to simplify Q-learning discrete value table into a new version that stores only one optimum action and its Q-value instead of storing every action's Q-value in each state. Exploration and contrast experiments show that our algorithm learns much faster than the original Q-learning without losing the ability to find a better policy in navigation task.
展开▼