Adaptive Sampling for Human-aware Path Planning in Dynamic Environments

Nowadays, robots are increasingly used in densely populated dynamic environments. Robots not only need to complete the navigation tasks quickly, but also need to take into account the human trajectories and the constraints of social rules. In order to avoid the robot going into the crowed areas and improve robot acceptance in the crowded public environment, we propose a human-aware motion planning algorithm that is based on sampling method. Firstly, human will be annoyed and stressed if robots disturb them during operation. To alleviate this uncomfort brought by the robot, we use probabilistic representations to build the Human Domain Zone (HDZ) of individual or crowd behaviors. Besides, we propose a sampling strategy that is capable of biasing the sampling in the area where human feel comfortable or crowd is sparse. Moreover, we put forward an evaluation function to select the optimum trajectory. This function can avoid robot falling into the crowded area through VDM which is used to model the relationship between human and robot. The proposed approach is verified with extensive experiments in simulated environments. The results show that our method has the promising performance in crowded environment. It can also generate a smooth path with higher success rate.