面向3C制造的多机器人协同敏捷装配共性理论与技术研究

It is an emergent yet challenging mission for Chinese manufacturing industry to achieve high speed/high precision robotic assembly performance for complex 3C (communication, computer and consumer electronics) components represented by integrated chips (ICs) and electrical boards. However, till date, there are still fundamental deficiencies in 3C assembly manufacturing, including high cost, low flexibility, weak reconfigurability and so on. To solve such a dilemma, this project aims at developing generic multi-robot-based intelligent assembly manufacturing theory and equipment for 3C component assembly processes by using both fused vision/force information and 3C assembly process technic knowledge. More precisely, the project will propose an environment recognition method based on fused multi-sensor information, and thereby develop an integrated scheme with real-time geometric assembly error identification, error compensation, obstacle avoidance and motion planning. Moreover, an assembly force/displacement collaboration control approach will be provided together with a self-organized pattern regulation method for multiple manipulator tips. By this means, a new interactive principle between the assembly process and its dynamic response will be revealed, and then a new dynamics analytical theory will be established for 3C multi-robot assembly systems. Furthermore, a self-organized pattern controller for the multi-robot assembly system will be developed as well. Finally, a multi-robot-based assembly equipment and the associated robotic machining software will be developed to accomplish the high speed/high precision robotic assembly exemplary application missions for electrical boards of servers. This project will provide an integrated multi-robot collaborative assembly solution scheme for complex 3C components, which is of both theoretic and application significance for promoting 3C intelligent manufacturing research.

以集成电路、电子板件等为代表的复杂3C零部件的机器人高速高精装配是我国3C制造业面临的严峻挑战。本项目针对3C装配存在的成本高、柔性差、可重构性不足等缺陷，以服务器电子板件装配为主要对象，将3C零部件装配技术与多机器人自组织协同理论方法结合起来，开展融合视觉、力等多源传感信息和装配工艺知识的群体机器人协同装配共性理论、技术与装备研究。研究内容包括多机器人装配的多源传感信息融合感知、几何误差补偿、实时避障与运动规划、装配力位协同控制、多机械手末端自组织构型调控等。旨在揭示群体装配机器人执行响应与工艺过程间的动态交互机理，提出3C零部件多机器人协同装配的新原理和新技术，研制具有关键技术突破的多装配机器人装备原型样机和软件系统，实现服务器零部件的高速高精装配应用示范。项目预期将为3C零部件装配过程提供多机器人协同装配解决方案，形成3C智能制造研究的新增长点。

计算机、通信和消费电子（3C）产品在人工智能、物联网、航空航天等行业应用广泛，其产品变化快、制造周期短，同时面临人工成本快速增长的困难。这些问题使得利用机器人完成多任务、小批量、高强度的高速高精装配任务，以实现敏捷、柔性和精益制造目标的需求日益迫切。本项目以此为背景，依次从环境感知和动力学建模、协同运动规划、自组织敏捷控制和原型系统研制四个层面开展研究，取得如下4个主要进展：.（1）建立了多机器人协同装配过程中机器人本体与操作工件间时变、耦合、切换的动力学模型，提出了面向多机器人协同装配场景的高精度轻量级的视觉感知模型，解决了传统多机器人协同装配模型中敏捷性不足的问题，较之通用视觉感知模型，平均视错误差降低16%，位姿识别时间缩减14%。.（2）发展了多机器人系统避碰避障策略，提出了基于强化学习的机器人变阻抗轨迹跟踪控制算法，设计了面向任务的机器人自主路径探索规划策略，阐明了机器人装配系统在不同装配任务驱动下的柔顺装配动作创成方法和切换机制，较之无学习的路径跟踪方法重复定位误差降低20%。.（3）揭示了通信网络对多机器人系统稳定性的影响机理，建立了切换拓扑下多机器人系统输入-状态稳定性判据。提出了多机器人协同装配中轨迹跟踪子任务和力维持子任务的状态空间划分方法，据此设计了多机器人分布式状态估计和协同操作策略，装配误差较之使用误差反馈的直接法降低20%。.（4）研制了多机器人协同装配系统原型机及服务器主板装配产线，开发了集成感知定位、避碰避障、协同控制的仿真软件，实现了2路和4路服务器CPU、内存条、散热器等3C零部件协同装配原理验证和产业示范，协同装配速度较之无机器人产线提高40%以上。.项目团队出版英文专著1部；在IEEE、ASME汇刊和Nature Communications等国际期刊发表SCI论文30篇；发表中文期刊和国际会议论文21篇；申请国家发明专利11项，其中授权4项；登记软件著作权3项；牵头获得湖北省自然科学奖1项。在学术交流方面，组织大型学术活动9项（国际性3项、全国性6项），作大会特邀报告10次。项目共培养博士后3人，博士生5人，硕士生11人。