仿人机器人反射式失稳应对理论与方法

In order to promote the humanoid robot's application in our daily lives, the theory and key technology should be broken through for humanoid robots to recovery from instability rapidly and reduce tumbling injury due to external disturbance or unexpected deviation from the environment. Up to now, previous researches in robot society focus on explicit mathematical model to obtain the balance recovery. However it is very difficult to model the actual disturbance and environment deviation into mathematical model. Whereas human body's reflexive recovery response capability can handle interference and environment deviation rapidly and universally to keep balance. At present, the research on instability reaction based on human motion principles has not been carried out, which is an urgent fundamental issue to be solved. In this project, we study the generation mechanism and process of human's reaction against instability, and model the humanoid robot's reaction model against instability. Then we explore the matching and mapping method of kinematics and dynamics between human and humanoid robot, reflexive recovery strategies to reaction against instability, the actively reflexive tumbling strategy for evitable falling that robot can't recovery due to large disturbance, and the evaluation for balance recovery robustness and tumbling injury. Finally, we will verify the proposed theory and methods via simulation and experiments. The contribution of this project is expected to propose a systemic theory of reflexive reaction against instability for humanoid robots to enhance their environmental adaptability.

解决仿人机器人在遭受外界干扰或地面环境突变而失去平衡（失稳）时快速应对从而恢复平衡、降低跌倒损害的基础理论和关键技术，是推进仿人机器人走向广泛应用的重要前提。目前国内外相关研究都局限于通过明确的数理模型求解，而实际的干扰和环境变化难以用特定的数理模型表达。然而人体无需数理模型，能快速而普适地反射式应对，克服干扰和环境变化恢复平衡。国际上基于人体反射运动的失稳应对研究还未开展，是一个亟待解决的基础科学问题。 本项目研究和借鉴人体失稳应对的机理、过程，建立仿人机器人反射式失稳应对模型；研究人体与仿人机器人运动学、动力学匹配与映射方法，仿人机器人失稳的反射式平衡恢复策略，以及大干扰下不能恢复平衡时的反射式主动跌倒方法；仿人机器人平衡恢复鲁棒性及跌倒损害评价等核心问题，并进行仿真和实验验证研究。项目目标是提出仿人机器人基于人体反射运动的失稳应对理论，为提升仿人机器人环境适应性提供一套理论和方法。

本项目围绕仿人机器人环境适应的前沿基础理论与方法，研究了人体失稳时的反射式平衡恢复和跌倒保护运动规律，建立了仿人机器人反射式运动模型，提出了一整套基于人体反射式运动规律的仿人机器人失稳应对理论与方法，并在已研制的“摔滚走爬”多模态运动仿人机器人“汇童”上进行了验证。首先，建立简化的人体模型，利用光学运动动捕捉系统与地面反力测量系统获取并分析了人体运动数据，得到了人体失稳下的反射式应对规律，建立了仿人机器人反射式运动模型。其次，基于仿人机器人反射式运动模型，研究了仿人机器人反射式平衡控制方法，提出了仿人机器人平衡恢复策略。研究了仿人平衡恢复鲁棒性评价方法，提取了表征机器人平衡恢复能力的参数体系，提出了以地面反作用力总冲量、冲量分布函数为评价指标的仿人机器人跌倒损害评价方法。然后，研究了仿人机器人与环境碰撞冲击建模，提出了仿人机器人着地姿态及其冲击力预估方法。基于三阶倒立摆模型与倒立摆飞轮模型，构建了仿人机器人反射式跌倒应对数据库。基于支持向量机与机器学习算法，提出了仿人机器人跌倒运动模式触发与生成方法。最后，研究了仿人机器人动力学仿真建模方法，建立了动力学仿真环境，进行了失稳平衡恢复和跌倒应对的仿真。并对仿人机器人实验平台进行了改进并进行了实验研究，验证了提出的仿人机器人失稳平衡恢复和跌倒应对等理论和方法的有效性。