Iterative Dynamic Waypoint Navigationによる複数移動障害物の回避軌道計画とその計算効率化

抄録

For providing services where environments exist multiple humans or obstacles, robots need to move with considering multiple humans simultaneously, react quickly to the behavioral changes of pedestrians, choose acceleration/deceleration appropriately, and predict human’s behavioral changes. We thus propose a multiple moving obstacles avoiding method, called Iterative Dynamic Waypoint Navigation (IDWN). The method consists of (i) interference estimation, to predict future interference with a pedestrian, (ii) cluster estimation, to avoid pedestrians as a cluster who cannot be avoided without entering their personal space, (iii) path generator, to calculate waypoints of default speed, acceleration and deceleration paths, (iv) iterative search, to repeat the above three functions for all pedestrians in the environment, and (v) cost calculation, to select the lowest cost trajectory. For reducing the calculation cost, some heuristics, omitting candidates which theoretically increase the cost, are introduced. We implemented the method into a simulator and conducted evaluation experiment by comparing with several conventional methods. As the result of the experiment, we verified that IDWN worked well in versatile dynamic environments with multiple moving obstacles with less computation time.