基于WGAN-GP的搜索式路径规划算法

为了提升搜索式路径规划算法在C字型障碍中的探索效率，提出了一种基于对抗生成网络的A*算法。首先使用训练更为稳定的梯度惩罚Wasserstein对抗生成网络（WGAN-GP）生成存在可行路径的感兴趣区域；然后使用A*算法优先探索该区域，使得路径规划能够被有效引导；最终形成一条连续的路径。经过实验仿真验证，其相较于传统A*算法节约了31%的规划时间、减少了22.84%的探索空间，提升了路径规划算法的效率。实验结果表明，改进的A*算法具有较高的探索效率，能够更好地应用于机器人路径规划中。

Search-based path planning algorithm based on WGAN-GP

In order to improve the exploration efficiency of the search-type path planning algorithm in the C-shaped obstacles, this paper proposed an A* algorithm based on the generative adversarial network. Firstly, the method used a more stable Wasserstein generative adversarial network with gradient penalty(WGAN-GP) to generate regions of interest with a feasible path. Then the algorithm used the A* algorithm to preferentially explore the regions, enabling it to efficiently guide path planning. Finally, it formed a continuous path. According to the experimental simulation verification, compared with the traditional A* algorithm, it saved 31% less planning time and 22.84% less exploration space, and improved the efficiency of the path planning algorithm. Experimental results show that the improved A* algorithm has high exploration efficiency, and it can achieve better application in robot path planning.