基于行人间交互意图检测的服务机器人导航

Service robot navigation based on interaction intention detection between pedestrians

针对人与机器人共存环境中的机器人导航问题,提出一种基于行人间交互意图检测的服务机器人导航方法.首先,融合RGBD(红-绿-蓝-深度)信息和激光信息获取准确的行人运动轨迹,从行人轨迹中提取交互意图特征,基于贝叶斯理论检测出行人间是否存在交互意图;然后,将存在交互意图的行人之间区域设置为交互区域;最后,在地图中的交互区域设置临时障碍物,使机器人路径规划避开交互区域.实验结果证明了所提方法的有效性.

To address the problem of service robot navigation in environment coexisting with humans,a robot navigation method with interaction intention detection between pedestrians was proposed.Firstly,the accurate trajectories of pedestrians were obtained by fusing the RGBD（red-green-bluedepth）and laser information.The interaction intention features were extracted from the trajectories,and interaction intentions between pedestrians were detected based on Bayes theory.Then,regions between two pedestrians were set to be interaction regions if they had intentions of interaction.Finally,temporary obstacles would be set on the map to force the robot to build a path that could avoid the regions.The effectiveness of the proposed method was verified by experimental results.