基于状态估计的服务机器人主动感知系统的研究

Research on Active Sensing System of Service Robot Based on State Estimation

服务机器人在人机交互过程中,一个重要问题是如何在未知环境下确定人的位姿并能够快速准确的运动到人的面前。针对这一问题,文章设计了基于激光传感器的服务机器人主动式感知系统。在本系统中,服务机器人在没有先验信息的条件下,可以通过运动过程中检测、更新环境中的人物位姿来降低目标位置的不确定性,并且实时规划可以到达该人物正前方的最优路径。系统采用基于Adaboost的模板估计算法来估计人物位姿,采用基于不确定性预测的多目标优化算法来实现运动决策优化。在机器人轮椅平台上实现了该系统,并对系统的性能进行了评估。

In the interaction process between service robot and people, it is a significant problem for the robot to quickly and actively localize the position of people and find an optimal trajectory to get close to people. For this problem, this paper proposes the active sensing of service robot based on the laser sensor, which can endow the robot the ability of real time detecting, calculating people＇ s pose and computing the optimal path to the people. The people pose calculating module adopts the Adaboost based pattern learning algorithm and the action decision making module utilizes the uncertainty reasoning based multi-goal optimization algorithm. The whole system is implemented on robotic wheelchair as well as evaluations of the performance are achieved, too.