Motion planner of mobile robots which avoid moving human obstacles on the basis of stochastic prediction

In this paper, a trajectory planning method by which autonomous mobile robots accomplish their tasks avoiding human obstacles with uncertain motions is proposed. Human motions in the near future are predicted by a motion predictor using a stochastic process model as probability maps of existence of obstacles. On the basis of these maps, time and magnitude of danger of collision are estimated Robot trajectories are determined so that a function evaluating planned trajectories becomes optimal. The characteristics of this method are that it does not need any heuristics for strategy of avoidance, and that the two problems of motion prediction and of motion determination are distinguished. Simulations were performed supposing that a man and a mobile robot coexisted and moved in a room. The results revealed that robots can determine suitable trajectories to the goals avoiding obstacles even if human motions dynamically change.