基于人臂运动机理的机械臂运动控制理论与实验研究

As the application of robots is getting closer to human, higher requirements to robots are put forward. Human-like motion is one of the important issues, especially to the humanoid service, advanced industrial robots and assistive robots. The first two kinds of robots will work in human environment and directly interact with human. Human-like motion can improve human-robot interaction and increase the efficiency. As to assistive robots, human-like features will greatly improve the treatment outcomes. In this project, human-like motion control for robotic arms is studied to make the robotic arms perform manipulation tasks in human-like manner autonomously. Firstly, human arms are taken as the research object. Its motion process is researched and the movement rules are explored. By analyzing the motion states in different moments, a human arm motion model is determined. Secondly, on the basis of those, the redundancy resolution of human arm in different tasks, the selecting mechanism of the relative positions between shoulder and object, the security of the man-machine interaction and the mechanism of the winning strategy in dual arm competition are studied respectively. Then, new planning algorithms are proposed for motion planning of human-like movements based on the Rapidly-Exploring Random Trees (RRT) and the evaluation systems of robotic arms for accomplishing manipulation tasks are built. Last, the research results are verified through the experiments of the two applications of humanoid service robots and advanced industrial robots. The innovations of this project are as follows: 1) New thought and new method. In the traditional researches of robotic arms, the objects of study are always static, but this project does the research on the human arm motion process. 2) New human performance measures and new algorithms for human-like motion control. In this project, new hypotheses, models and human performance measures are proposed to study the underlying mechanisms of human arm movements. Furthermore, new motion planning algorithms are proposed to make the path of the end-effector and the arm postures of robotic arms resemble human arm motions. 3) The new concept of dual arm competition. Inspired by Tai chi push hands, this project proposes a new model of dual arm operation named "dual arm competition", to study the associated new problems.

随着机器人越来越接近人类，对机器人提出了更高的要求。仿人运动是其中重要的一项，既可以给人类使用者营造安全舒适的合作环境，也可以进一步提高人机合作的效率。本项目针对拟人机械臂的仿人运动控制问题，以人臂为研究对象，首先对人臂运动过程进行研究，探索人臂运动规律，分析运动过程中不同时刻下的人臂运动状态，确定人臂运动模型。在此基础上，对不同任务下的冗余度分解机理，肩部和操作对象之间相对位置的选择机理，人机交互安全性原则和双臂对抗中的制胜策略机理展开研究。同时基于快速扩展随机树算法专门针对仿人运动规划这一应用背景，建立新的规划算法和完整的拟人机械臂仿人运动的综合评价体系。最后，针对"友好、协作"特点，从服务型机器人和新型工业机器人两方面进行实验研究，并验证理论结果的准确性。项目研究成果对阐明人臂运动内在机理，揭示人臂运动规律具有重要意义，为拟人臂的仿人运动控制及应用奠定理论基础。

随着机器人越来越接近人类，对机器人提出了更高的要求。仿人运动是其中重要的一项，既可以给人类使用者营造安全舒适的合作环境，也可以进一步提高人机合作的效率。本项目主要研究基于人臂运动机理的拟人机械臂的运动规划和控制问题，目标是使机械臂以拟人的运动方式与人发生交互行为。首先基于大量的人臂运动捕捉实验结果，建立了拟人化臂姿指标和运动阶段划分法来解析人臂的运动机理。其中拟人化指标基于人臂最小势能和最大舒适度原理，通过优化该指标值确定拟人臂姿。而运动划分则基于距离判据和姿态判据，实现完整运动的分割。其次，基于运动划分的方法提出一种分级规划策略来模拟人臂运动过程中运动状态的变化，将它与基于贝叶斯网络的人工智能决策手段相结合，建立了一种新的拟人运动规划方法，并通过耦合神经网络反解出拟人臂的运动关节角来实现机械臂的运动控制。同时，基于仿人运动的运动参数建立了一种情感模型，通过该模型识别人臂运动表达出来的情感并把它用于规划机器人的运动，让机器人在完成任务的同时也展现出不同的情感。接着，建立了多种仿人程度评价指标和其他机器人性能指标，实现对机器人仿人能力和综合性能的评价，用于对机器人的进一步改进。而后，针对机器人的安全性问题进行了综述，提出基于关节空间快速组合搜索的安全性避障算法，并针对协作问题提出一种基于误差空间分析的协作运动规划方法。此外，提出了基于重力补偿方法和预设性能函数方法的机械臂运动控制方法，实现更好的轨迹跟踪控制和更高的控制精度，以满足拟人臂的仿人运动控制需求。最后，开展运动捕捉实验和拟人臂仿人运动控制实验，验证了新方法的正确性和有效性。本项目的研究成果揭示了人臂的内在运动机理，提出的新方法能够应用到人形服务机器人、新型工业机器人、医疗康复机器人的拟人臂的运动控制中，提高它们的仿人能力和与人交互的能力，具有重要的现实意义。