Abstract: The authors present a body state estimator for a quasi-passive bipedal robot with dynamic gaits, to estimate the key performance parameters of the robot during stable walking motion. Based on the traditional inertial measurement unit, this approach integrates accurate pitch and roll measurement result to correct the impact for acceleration by gravity, in order to improve the performance of the body state estimator. The method implements the estimation procedure off-line, using data extracted from various repeated walking motion of the robot and evaluates its performance with reference to a visual ground-truth measurement system. Satisfactory results are obtained by using body state estimator to analyze both robot walking and human walking. The results can be used as feedback to improve the stability of the walker.

Key words: quasi-passive dynamic walking, bipedal robot, body state estimation, visual ground-truth measurement system

摘要： 针对现有半被动双足机器人缺乏对环境的感知且行走稳定性差等问题，研发了一款位姿估算器。在传统的惯性测量单元的基础上进行了改进，用精确的俯仰和翻滚测量结果修正了重力加速度对加速度计产生的误差，从而提高估算器的性能。利用机器人多次实际实验行走所获得的离线数据得出行走参数，并且和地面实际测量系统的测量结果进行了比较。实验证明所研发的位姿估算器具有令人满意的测量精度，可准确的测量机器人行走时的关键性能参数。测量结果作为反馈可有效提高半被动机器人的运动稳定性。

关键词: 半被动动态行走, 双足机器人, 位姿估算, 地面实际测量系统