分布式机器视觉的RTPS中间件设计与实现研究

Design and implementation of the RTPS middleware of distributed machine vision systems

为了满足网络化制造中分布式机器视觉（DMV）对各种数据流的不同通信性能要求，设计了基于实时发布-订阅（HTPS）协议的中间件。通过定义中间件工作过程中的相关控制参数，对数据传输的可靠性、实时性以及网络带宽和内存等资源的使用进行了定量描述和优化。将设计的RTPS中间件应用到网络化制造中的DMV系统，进行了各种类型数据流的不同通信性能实验，包括视觉图像信息数据流，该数据流可以选择异步通信或者同步通信。对系统进行了建模和模拟实验，结果表明，与基于Client／Server的中间件相比，基于RTPS协议的中间件的延迟低，在数据包大小变化时的稳定性强，而且吞吐量很大。可见，所设计的RTPS中间件能保证DMV的实时性、可靠性、动态适应性，以及多节点对多节点的通信要求。

To meet the communication performance requirements of a network-based manufacturing system＇s distributed machine vision （DM-V） in data flows, the middleware was designed based on the research on the real-time publish-subscribe （RTPS） protocol. Control parameters were introduced to optimize the reliability and the real-time of data transmissions, as well as the use of the memory and the network bandwidth. The designed RTPS middleware was applied to a DMV system in the network-based manufacturing environment, and different communication experiments on various types of data streams were completed. In particular, the visual image information data streams could choose the type of asynchronous communication or synchronous communication. The experimental results demonstrate that the RTPS middleware is suitable for the timelines requirements for DMV systems.