在3C、家电等消费类行业中常见的弱刚度零部件（如复合材料部件、柔性电路板、塑料外壳等）具有易变形、弯折、损坏等特性，装配过程中面临着柔性装配状态描述困难、装配过程操作复杂、装配成功率低等问题，现有工业机器人装配算法主要面向刚性部件，且装配工艺相对固定，难以适应弱刚度部件装配作业需求，无法保证装配效率和装配质量。本项目结合新一代人工智能技术，拟开展弱刚度零部件柔性装配状态获取与过程描述，基于深度强化学习的插装、卡合、贴装等装配技能获取，装配过程状态多模信息评价体系建立及柔性装配工艺参数优化等研究内容，并搭建仿真与实验平台验证算法的有效性，实现基于技能增值的工业机器人弱刚度零部件柔性装配，提升装配成功率。本研究对机器人的运动灵巧性和平稳性以及执行任务的实时性均具有重要的理论研究意义，并为今后的产业化应用提供技术支撑，有效的推动装配行业自动化和智能化的发展。

Weak-rigidity components, such as composite components, flexible circuit boards, and plastic shells, are widely utilized in manufacture of consumer electronics. In the process of their automatic assembly, it is easy to be deformed, bent and damaged. There are some problems, such as insufficient description method of flexible assembly state, complex operation of assembly process and low success rate of assembly. Traditional high precision assembly algorithms of industrial robots are mainly for rigid components, and the assembly process is relatively fixed. It is difficult to adapt to the assembly requirements of components with weak-rigidity, and the assembly efficiency and quality cannot be guaranteed. Combining with new generation artificial intelligence technology, this project will solve problems of the weak stiffness components assembly. It intends to adopt multi-mode information state description method with deep learning to obtain the flexible assembly state of parts with weak-rigidity and describe the assembly process. The method of skill acquisition based on deep reinforcement learning is studied. It can realize skill acquisition of unit actions, such as interpolation, snap and paste, under the constraints of multi-mode information. Based on the multi-mode information evaluation system, the parameter optimization method of flexible assembly process is studied to improve the assembly success rate. The performance of proposed algorithms and models will be validated on the simulation and real platform. The research has important theoretical significance for the robot dexterity and stability and the real-time performance of the task execution. It provides technical supports for industrialization in the future and promotes the automation and intelligent development of the assembly industry effectively.