面向人机协同的机器人智能制造感知与作业理论研究

Intelligent manufacturing 2025 is an important developing strategy of our country. Due to the complexity of the production environment and the diversity of the operating targets， an industrial robot system needs to have characteristics of robust recognition and high-precision positioning via visual, quick and accurate operation ability and safe and compliant human-robot collaboration, which challenges the current robot theories and methods. Aiming at complex industrial production environment, this project proposes theories and methods of visual robust identification and high-precision positioning combinig with maximum correntropy Kalman filter and local feature matching; extracts the operating characteristics of target to form parametric model and studys the rapid and stable operation strategy; integrates advantages of robots’ high-precision positioning and human’s flexible operation ability to explore a quick and accurate path planning method for industrial robots; operates research on human-robot collaboration theories and methods of synergy compliant safety operation, solves low productivity problems of industrial robots’ soft and intelligent production, and realizes the application of industrial robots in high performance intelligent manufacturing. This project combines technologies of robotics with bionics, intends to break through the intelligent manufacturing theories and methods of robot in the complex industrial environment, promotes the application and development of industrial robots in the advanced manufacturing, and provides a basis for the leading of our country’s manufacture reforming. This project has significant economic and social values, and it is important for further promoting the level and strategic position of our countrys’ manufacturing industry.

智能制造2025是我国重要发展战略。制造过程中，环境复杂且作业目标多样，下一代机器人需要高精度定位、鲁棒识别、精准快速的操作能力和人机交互协同柔顺安全作业的特性，对现有的感知及控制理论和方法都提出了新的挑战。项目针对工业复杂生产环境，提出结合最大相关熵卡尔曼滤波与局部特征匹配的机器视觉鲁棒识别与高精度定位理论与方法；提取作业目标操作特征参数化模型并探究快速稳定操作策略；综合机器人定位精度高和人灵活作业能力强的优势，探索工业机器人快速精确路径规划方法；研究人机交互协同柔顺安全作业理论和方法，力图解决工业机器人智能柔性化生产能力低问题，实现机器人高性能智能制造应用。项目结合机器人学与仿生学技术，突破机器人的复杂工业环境智能感知与精准控制的理论和方法，促进其在智能制造中应用和发展，为引领我国制造方式的变革奠定基础，对进一步提升我国制造业的水平有重要意义。

项目针对工业复杂生产环境，研究机器视觉鲁棒识别与高精度定位理论与方法；综合机器人定位精度高和人灵活作业能力强的优势，探索工业机器人快速任务规划方法；研究人机交互协同柔顺安全作业理论和方法，为工业机器人智能柔性化生产能力低问题提供了解决途径。项目结合机器人学与仿生学技术，突破机器人的复杂工业环境智能感知与精准控制的理论和方法，促进其在智能制造中应用和发展，为引领我国制造方式的变革奠定基础，对进一步提升我国制造业的水平有重要意义。.项目提出了基于数字孪生系统的人机环交互动力学模型，揭示了人机多点接触交互机理；针对人机意图传递问题，完成了基于阻抗模型的机器人力跟随控制及人类柔顺行为学习，在任务层面上提出了基于场景图的机-环交互任务规划方法；针对非结构化环境下目标位姿估计的难题，提出了三角化几何特征约束与加权损失函数的关键点定位网络，实现了复杂环境下基于单目RGB图像的高精度位姿预测；针对五指手自由度高、抓取动作类别多的特点，提出了基于概率密度重采样的五指灵巧手抓取动作生成方法，将手指关节空间由20维度降低到5维度并在低维度中用核密度估计方法拟合手指关节角度的概率分布密度，为快速生成抓取路径提供基础。.本课题发表IEEE TRO、IEEE TIM、IEEE TII等高水平论文41篇，申请或授权发明专利20项，工作获得2019年度上海市青年科技杰出贡献奖1项，天津市科技进步二等奖1项，培养博士研究生17名，硕士研究生13名。