复合地面环境下双足机器人全/欠驱动混合的仿人行走稳定性控制

Although the great application requirements of biped robots in dangerous circumstances such as outdoors and disaster rescue in the near future, current pedestrian mode, gait planning methods and control strategies with separated full/under-actuated mode can not meet the demands of high stability and energy efficiency walking on the random composite ground. This project proposes the full/under-actuated hybrid walking mode based on the human gait characteristics. Focusing on the scientific problems of status perception and walking control of biped robots on the composite ground environment, we will study the following five aspects. The first is to extract features between the ground and the humanoid lower limb walking robot. The second is to study the dynamics of the ‘robot-composite ground’ coupling system. The third is to analyze the stability of the full/under-actuated hybrid walking mode when facing the disturbance of random composite ground. The fourth is to identify walking status with imitating the human behaviors. The last is to propose humanoid layered anti-disturbance control strategies and adaptive control algorithms for multi-tasks. Research results will reveal the influence rules of composite ground disturbance on the robot posture, provide the stability criterion for full/under-actuated hybrid walking mode and solve the key problems such as status perception and anti-disturbance recovery in the real-time control condition. At last, a prototype of the biped robot using full/under-actuated hybrid walking mode will be developed. The implementation of this project will improve the gait compliance and resistance ability under composite ground disturbance, increase the walking energy efficiency and provide the fundamental theory and technical support in practice for biped robot.

双足机器人在野外与救灾等场合有广泛的应用需求，而当前采用的全/欠驱动分离的步行模式、步态规划方法与控制策略尚不能满足随机复合地面干扰下的步行稳定性与行走能效要求。本项目基于人类步态特征，提出全/欠驱动混合行走方式，并围绕复合地面环境下的仿人步行状态感知与稳定步行控制这一科学问题，重点研究地面与人类下肢特征提取、机器人-复合地面耦合动力学特性、随机复合地面干扰下机器人全/欠驱动混合行走模式的稳定性能、模仿人类感知行为的步行状态辨识、面向多任务的仿人分层抗扰控制策略与自适应控制算法。预期成果：揭示复合地面干扰对机器人步行姿态的影响规律；提出适用于全/欠驱动混合行走模式的稳定性判据；解决能满足实时控制要求的步行状态感知与抗扰恢复等关键问题；形成全/欠驱动混合步行机器人样机。本项目的研究可增强双足机器人的步态柔顺性与对地面扰动的抵抗能力，提高行走能效，为其走向实用化提供基础理论和关键技术支撑。

双足机器人形态拟人，对非结构化环境具有良好的适应能力，且无需为其改造人类工作和生活的环境，得到了研究者的广泛重视。而当前采用的全/欠驱动分离的步行模式、步态规划方法与控制策略尚不能满足随机复合地面干扰下的步行稳定性与行走能效要求。本研究基于人类步态特征，提出全/欠驱动混合行走方式。.主要研究内容与成果包括：1）采用解耦建模的方式分别建立前向平面和侧向平面“机器人-柔性地面”耦合动力学模型；在此基础上提出水平地面上欠驱动双足步行控制算法和变高度台阶上欠驱动双足步行控制算法。2）基于强化学习设计欠驱动双足机器人髋关节刚度控制策略；针对全驱动双足机器人的三维步行任务，借鉴生物的主动节律运动，设计基于强化学习的周期步行控制器。3）提出基于多策略集成的双足步行分层控制算法，形成包含步态规划层、姿态调控层和伺服跟踪层的稳定步行控制结构。4）最后，研制欠驱动双足步行试验平台，验证所提出控制算法有效性。.本项目的研究增强了双足机器人的步态柔顺性与对复合地面扰动的抵抗能力，提高了行走能效，为其走向实用化提供基础理论和关键技术支撑。