A Generative Adversarial Network Based Motion Planning Framework for Mobile Robots in Dynamic Human-Robot Integration Environments

In the human-robot integration environment, efficient and safe navigation is of great significance for mobile service robots. At present, human-robot integration environment is highly uncertain and dynamic, which brings new challenges to motion planning. In order to solve this problem, this paper proposes a dynamic obstacle avoidance strategy based on imitation learning in a Generative Adversarial Network (GAN) framework. When the robot detects a pedestrian around it, it generates an active obstacle avoidance point that maintains an appropriate distance from the pedestrian according to the pedestrian pose and the global path planned by the A* algorithm as a sub-goal to guide the robot for motion planning. In the experiment, the performance of the algorithm is evaluated by the number of entering the pedestrian person space, the time cost and the trajectory length. Compared with the Dynamic Window Approach (DWA) and Proactive Social Motion Model (PSMM) algorithms, the experimental results show that our proposed algorithm has better performance than the other two algorithms in the human-robot integration environment.