基于改进Q-learning算法的移动机器人路径优化

Improved Q-learning Algorithm for Mobile Robot Path Optimization

针对多障碍复杂环境下移动机器人路径规划问题,提出了一种基于萤火虫算法的Q-learning算法。在求解算法中,为提高算法的收敛速度,使用萤火虫算法初始化Q-learning框架下Q值;为平衡算法搜索的随机性和目的性,结合贪婪搜索与玻尔兹曼搜索,设计了混合选择策略,使得算法可以动态选择搜索策略,并通过仿真实验验证所提算法的有效性。仿真实验结果表明所提算法在计算时间和路径平滑度等指标上优于Q-learning算法和Sarsa算法。

Aiming at the path planning problem of mobile robot in multi-obstacle complex environment, a Q-learning algorithm based on firefly algorithm is proposed. In the solution algorithm, in order to improve the convergence speed of the algorithm, the firefly algorithm is used to initialize the Q value under the Q-learning framework;In order to balance the randomness and purpose of search, combined with greedy search and Boltzmann search, a hybrid selection strategy is designed, so that the algorithm can dynamically select the search strategy. The effectiveness of the proposed algorithm is verified by simulation experiments. Simulation results show that the proposed algorithm is better than Q-learning algorithm and Sarsa algorithm in terms of calculation time and path smoothness.