基于人臂运动特性的柔性双仿人臂点位约束作业规划与关节刚度控制

The joint structure of the 7-DOF humanoid manipulator is in accordance with the physiological characteristics of human arm, so it has large workspace. However, as to the complex tasks in actual application, if two humanoid manipulators of the humanoid robot can cooperate with each other mutually, it would bring a greater advantage from the mechanism performance. Based on the arm motion characteristic of human being, in order to address the problems of dynamic/static obstacle avoidance, low energy efficiency and the position and orientation disturbance of the humanoid body in point-to-position constraint task for the dual humanoid manipulators, the trajectory planning of the dual humanoid manipulators for coordinated operation is transformed into the trajectory parameterization optimization problem. Based on the dual optimization criterion of energy efficiency and body position & orientation disturbance, multi-objective evolutionary algorithms(MOEAs) is adopted to optimize th trajectory and the Pareto non-dominated solution sets can be obtained. Then a optimal compromised solution is selected from the Pareto set and considered as "knowledge" information for the coordinated operation of dual humanoid manipulators, which can make the trajectory meet the physical constraints and point-to-position task constraints condition. In addition, in order to solve the security interaction problems between the robot and the environment, illumined by the motion characteristic that the joint stiffness of the human arm can be adjusted, an adjustable stiffness actuator-based bionic joint that possesses rigidity and flexibility is designed. Moreover, a composite control method that combines feed-forward and neuronal networks PID feedback is proposed to achieve the joint with adjustable stiffness control characteristics. In a word, the project study can solve several key technical problems and meet the requirements for the point-to-position constraint task of the dual humanoid manipulators, which has an important practical significance on promoting the application of humanoid robots.

仿人机器人7-DOF手臂结构符合人臂生理特点，具有较大工作空间，但在实际复杂任务执行中，若双臂能相互协调作业其机构性能将具有更大优势.本项目基于人臂运动特性，针对双臂在点位约束任务中存有动态/静态障碍、作业能耗大及对本体位姿扰动问题，将双臂作业轨迹规划问题转化为动作轨迹参数化优选问题，以能量利用率与对本体扰动优化双目标准则，采用多目标进化算法优选作业轨迹的Pareto非支配解集，并选取最优解作为动作学习的“知识”信息，使双臂在机构与任务约束范围内模拟人臂运动特性作业；另外针对双臂作业与环境间柔顺交互与安全问题，借鉴人臂关节"刚度"可调特性，设计一种基于可变刚度执行器的刚柔一体仿生关节，研究基于前馈与神经网络PID反馈相结合的关节刚度复合控制方法，实现刚度可调控制的良好特性. 通过本项目研究可解决双臂点位约束作业中关键技术问题，满足双臂协调作业任务需求，这为推动仿人机器人实际应用具有重要意义.

仿人机器人7-DOF手臂结构符合人臂生理特点，具有较大工作空间，但在实际复杂任务执行中，若双臂能相互协调作业，其机构性将具有更大优势。目前双臂运动控制方法研究已有一些成果，但在点位约束操作中仍存在作业动态障碍、静态障碍约束，与环境间柔顺交互以及安全等问题。针对双仿人臂在点位约束作业过程中存在的问题与挑战展开本课题的研究并取得如下成果：. 首先，分析人体手臂生理特点及结构，借鉴人体上肢双臂生理特点及关节结构，构建柔性双仿人臂系统关节结构模型。其次，提出冗余仿人臂基于最小加速度的轨迹优化, 这一定程度上能减少仿人臂作业对本体的反作用扰动； 同时，提出仿人机械臂飞行球体作业最优击球构型的求取方法，解决了仿人臂在任务操作过程中的运动学构型问题。接着，由于仿人臂点位约束作业动作轨迹规划问题存在多个优化目标，研究提出一种多目标精英反馈教学优化算法(MEFTO)，这可以有效解决实际工程中的多目标优化问题。在此基础上，基于人臂作业的运动特性，提出冗余仿人臂一种点位约束作业的多目标轨迹规划方法；同时，研究双仿人臂点位约束动作规划的数学模型及提出相应双臂协调规划策略，通过双臂叠杯子仿真实验也验证了该双臂点位约束方法的可行性。再有，提出一种径向基神经网络的分级学习方法，该学习方法优化变量参数少(3个)，具有强实用性，这为机器人关节控制器设计中基于神经网络的模型辨识提供有效方案。此外，设计一种基于弹簧片的刚柔一体变刚度关节，实现关节刚度可调的特性。最后分别搭建自行研制的串联弹性驱动器试验平台与基于弹簧片的变刚度关节试验平台，完成相应软件开发进行验证。. 在本项目资助下，发表6篇SCI/EI收录学术论文，其中SCI期刊检索1篇，EI检索5篇，包含其中的一级学报刊物1篇；授权国家发明专利2项，公示发明专利2项, 授权实用新型专利6项，正在受理实用新型专利3项；获批软件著作权1项，指导在读研究生7名。