中文摘要：

多轴超精密同步运动控制系统采用并行多处理器架构,提出基于分布式共享存储思路的数据交互机制,应对该控制系统中的高效数据采样与传输。设计含自定义总线的多总线运动控制网络,在分布式物理内存上实现统一编址的共享内存模型,通过基于硬件实现的消息传递服务刷新共享数据而维护其一致性,减小共享数据访问开销;将数据交互过程与实时控制算法呈流水线进行,优化并行计算周期时隙分配,最大化并行处理器使用效率;制定数据交互过程的精密同步策略,实现激光干涉测量位置数据同步采样。试验表明,数据交互机制性能不受并行处理器数目增多的影响,能达到稳定高带宽和精确数据传输延迟;同步策略能实现亚纳秒级别的位置数据采样同步误差,且该误差随信号传输距离稳定分布。应用实例证明了本研究内容的有效性。

英文摘要：

Multi-axis ultra-precise synchronous motion control system needs to adopt parallel multiprocessors architecture. A data interchange mechanism based on distributed shared memory to realize high-efficiency data sampling and transmission in this type of control system is proposed. Multi-bus motion control network involving user-defined internal bus is designed, which constructs globally unified addressing memory model based on ~listributed local memory. Hardware-implemented message passing service refreshes shared data to ensure data consistency, which actually minimizes memory access time. Pipelined execution of data interchange and real-time control optimizes the slot allocation in parallel computing cycle, which maximizes the efficiency of parallel-processors. Precise synchronization strategy in data interchange realizes synchronous sampling of laser measuring position datum. Experimental results show the performances of data interchange mechanism not affected by increasing number of processors can realize stable bandwidth and precise transmission latency. Sub-nanosecond scaled synchronous position sampling can be implemented, and synchronization error has stable distribution characteristics towards transmission distance. The validity of this research has been proved in actual application.