论文面向协作机器人控制架构的灵活性、实时性和柔顺性等需求，围绕机器人软件架构、实时系统和控制算法三方面核心内容，开展了协作机器人分布式实时控制架构设计、分布式控制器实时性能优化和评估、柔顺运动控制算法设计以及双臂协作机器人同步控制等方面的研究，具体内容如下：针对开放化、分布式、实时计算的协作机器人控制系统架构设计难题，提出了数据流的软件设计思路，构建了基于ROS2的分布式实时控制系统架构(ROS2-ADRCS)，完成了工业以太网EtherCAT总线、底层运动控制和上层规划算法在一个系统上的集成，支持基于组件的软件工程，并发和分布式执行程序，实现了机器人系统分布式分频控制，提升了协作机器人控制系统的灵活性、扩展性和交互性。构建了全抢占和事件驱动的系统内核，提出了一种结合公平调度和先进先出调度策略的协作机器人控制系统实时性优化方法，利用硬件时钟信号多级传递，采用信号量和锁存的方式实现了分布式系统数据同步发送和接收，提高了机器人数据传输的及时性和可靠性。并从吞吐量、运行频率、服务质量、订阅节点数量等多个角度对ROS1和ROS2实时性能进行了系统性评估，优化后的控制系统表现出了有效的实时性，提升了ROS2框架下机器人系统实时计算的能力。建立了具有关节柔性的双臂协作机器人模型，通过简化的弹簧阻尼系统进行连接，为控制算法的分析和验证提供了模型基础，定性和定量评估了刚性机器人和具有关节柔性机器人非线性动力学不同项所产生的力矩分量，相对于刚性机器人模型，具有关节柔性的机器人模型控制更加困难，静态和动态误差更大，在关节处容易产生波动，提出了基于模型的全状态控制算法，实现了位置、阻抗和导纳控制策略，相对于具有重力补偿的比例微分控制算法，所提出算法的位置、速度和力矩收敛更快，动态和稳态误差更小，能够有效抑制振动，提升了机器人动态性能。基于以上理论体系与关键技术，形成了分布式实时的软件系统，利用x86-64架构，实现了硬件资源的灵活扩展，搭建了基于PC的硬件平台，形成了独立的控制器，实现了双臂协作机器人3C元器件和CPU的装配。本论文研究工作对开发开放化、分布式和实时的协作机器人控制器具有一定的学术和工程应用价值。

Focusing on the demands for flexibility, real-time performance, and compliance in the design of collaborative robotic control architectures, this thesis investigates three key aspects: software architecture for robotics, real-time systems, and control algorithms. The research explores the development of distributed real-time control architectures for collaborative robots, the optimization and evaluation of the re-al-time performance of distributed controllers, the design of compliant motion con-trol algorithms, and synchronized control for dual-arm collaborative robots. The specific details are as follows:A data-flow-oriented software design approach is proposed to tackle the chal-lenges in designing open, distributed, and real-time control systems for collaborative robots, resulting in an architecture of distributed real-time control systems based on ROS2 （ROS2-ADRCS）. This architecture enables the integration of EtherCAT in-dustrial Ethernet bus, low-level motion control, and high-level planning algorithms into a single system. The approach supports component-based software engineering, concurrent and distributed program execution, and distributed frequency-divided control for robotic systems, thus enhancing system flexibility, scalability, and inter-activity.A fully preemptive and event-driven system kernel is developed, alongside an optimization method that merges fair scheduling and first-in-first-out scheduling strategies to augment the real-time performance of collaborative robot control sys-tems. Using multi-level transfer of hardware clock signals, and employing sema-phore and memory locking methods, the synchronization of data publishing and re-ceiving in distributed systems is realized, thus enhancing the timeliness and reliabil-ity of robot data transfer. A comprehensive evaluation of the real-time performance of ROS1 and ROS2 is conducted from various perspectives, including throughput, operation frequency, quality of service, and subscription node count. The optimized control system showcases effective real-time performance, bolstering the real-time computational capabilities of robotic systems within the ROS2 framework.A model of a dual-arm collaborative robot with joint compliance is established, connected via a simplified spring-damping system, thus providing a foundation for the analysis and validation of control algorithms. Qualitative and quantitative evalu-ations of the torque components produced by different terms in the nonlinear dy-namics of rigid robots and robots with joint compliance are performed. In compari-son to rigid robot models, the control of robot models with joint compliance presents a more significant challenge, characterized by larger static and dynamic errors and an increased propensity for fluctuations at the joints. A Model-Based Full-State Control Algorithm （MBFSC） is proposed, implementing consistent position, imped-ance, and admittance control strategies. Compared to gravity-compensated propor-tional-derivative control algorithms, the proposed method demonstrates faster con-vergence in position, velocity, and torque, coupled with reduced dynamic and steady-state errors. This effectively suppresses robot vibrations, thus enhancing the dynamic performance of the robot.Building upon the aforementioned theoretical framework and key technologies, a distributed real-time software system is developed, utilizing the x86-64 architec-ture to enable flexible hardware resource expansion. A PC-based hardware platform is established, generating an independent controller and facilitating the assembly of 3C components and CPUs for dual-arm collaborative robots. The research presented in this dissertation holds academic and practical value for developing open, distrib-uted, and real-time collaborative robot controllers.