人机协作环境下网络化机器人高性能控制关键技术研究

For the accuracy of robots and flexibility of humans, “human-robot collaboration” has been significantly oriented the development of robotics. Meanwhile, in the integration development of “Made in China 2025” and “Internet +”, networked robot has also been developing rapidly because of its advantages of data processing and sharing. Therefore, the networked human-machine collaborative robot which combines the two advantages will have great potential for application. Compared with the traditional application, the work environment of networked human-machine collaborative robot is more complex, which puts forward higher requirements on robot control quality. In view of this, by system modeling, error analysis and correction, controller design and experimental verification, the challenges facing the high performance control of the networked human-machine collaborative robot are systematically studied by this project. Since the traditional high performance control cannot be applied to the complex network environment and is sensitive to the unknown disturbances and also hard to real time implementation, several schemes including finite time robust adaptive control, robust predefined performance adaptive control, fast model predictive control, and predefined-performance-model-predictive compound control are investigated in this research. On this basis, a new practical high performance control design method is finally developed for the networked robots applied to complicated scenario such as human-robot collaboration, which can meet the high quality control needs with real-time, fast, precise characteristics for multiple state variables.

近年来，结合机器人的准确度和人的灵活性，“人机协作”成为机器人领域重要发展方向。与此同时，在“中国制造2025”与“互联网+”融合推进过程中，网络化机器人因其数据处理和共享优势也得到了快速发展。鉴于此，结合两方优势的网络化人机协作机器人将具有巨大应用潜力。相较传统应用，人机协作作业场景更加复杂，对机器人控制品质提出了更高的要求。对此，本项目以满足网络化机器人应用于人机协作等复杂作业场景时面临的多状态变量的实时、快速、精准高品质控制需求为目标，通过系统建模、误差分析与修正、控制设计和实验验证，系统研究网络化机器人高性能控制面临的挑战。为解决传统高性能控制难以适用复杂网络环境、对未知扰动敏感、实时性差等应用受限难题，提出了有限时间鲁棒自适应控制、预定性能鲁棒自适应控制、快速模型预测控制、预定性能-模型预测复合控制等方法。本项目的研究可为网络化人机协作机器人提供一套实用高性能控制综合设计方法。

本项目针对机器人系统在执行人机协作作业中面临的环境感知和多约束控制需求，开展了深入研究，提出了一些列实用高性能控制方法和机器人智能感知方法。该课题为机器人智能控制与应用领域研究的热点之一，综合了机器人、控制论、最优化理论、人工智能、大数据等领域研究，研究难度大。首先，提出了一种融合人体电生理信号与力传感器的新型柔顺控制方法，通过人机协同搬运任务验证了有效性。针对传统描述性能控制方法应用于实际机器人系统时，易受到未知扰动影响从而发生控制奇异，影响控制稳定性问题，提出了多传感器数据融合方法和鲁棒类描述性能控制方法，在提高控制性能的同时，提高了控制器的可实施性。针对传统有限时间控制、描述性能控制等方法得到的性能约束与机器人初始状态有关，不利用工程快速部署实践问题，提出了一种简化的预定性能控制方法。同时，也应用模型预测控制策略、控制障碍函数方法，从多角度提出机器人鲁棒多约束控制策略，进一步确保了机器人控制性能。项目搭建了脊柱手术遥操作机器人系统，验证了所提控制算法的有效性。最后，在高性能控制基础上，项目组还开展了机器人智能感知关键技术研究，提出了机器人系统人机交互智能感知与规划控制方法，进一步提高了机器人智能化水平。.依托该项目，项目组共发表SCI期刊论文13篇，EI会议论文9篇，出版中文专著1部，授权发明专利1项，获吴文俊人工智能自然科学二等奖1项。