Real-time Distributed Control: Optimizing Mechanical Performance using Adaptive Computing and Communication Techniques

PIS：D.M.Tilbury和N.R.Soparkar，密歇根大学提案编号：9977179实时分布式控制：使用自适应计算和通信技术优化机械性能本研究针对从机器人加工到自动车辆控制的应用领域的趋势，即从单一的、集中的控制算法转向分散的、分布式的实时控制实施。这种控制系统的整体性能不仅取决于所使用的控制算法，还取决于实现这些算法的计算硬件和软件模块。这个项目的目标是开发一个框架，在其中可以设计、分析和优化分布式控制系统。将考虑不同的控制体系结构(将处理任务分配给硬件模块)，并将根据基础计算基础设施(网络、处理器等)来表征每个体系结构中固有的延迟。分析了各种延时对系统机械性能的影响。这些信息将用于通过自适应计算和通信来改善机械性能，即将优先级分配给对控制系统的机械性能影响最大的计算或通信任务。机械工程和计算机科学在与工业相关的实时分布式控制应用领域的这种整合有望在这两个领域产生基本和实用的结果，并应立即适用于现实世界的问题。

PIs: D. M. Tilbury and N. R. Soparkar, University of MichiganProposal Number: 9977179Real-Time Distributed Control: Optimizing Mechanical Performance using Adaptive Computing and Communication TechniquesThis research addresses the trend in application areas that range from robotic machining to autonomous vehicle control of moving away from monolithic, centralized control algorithms towards decentralized, distributed real-time control implementations. The overall performance of such control systems depends not only on the control algorithms which are used, but also on the computing hardware and software modules which realize them. The goal of this project is to develop a framework in which distributed control systems can be designed, analyzed, and optimized. Different control architectures (allocation of processing tasks to hardware modules) will be considered, and the delays inherent in each architecture will be characterized based on the underlying computing infrastructure (networks, processors, etc). The effect of each type of delay on the mechanical performance of the system will be analyzed. This information will be used to improve the mechanical performance using adaptive computing and communication, that is, assigning priority to the computational or communication tasks which have the largest impact on the mechanical performance of the control system. This integration of mechanical engineering and computer science in the industrially-relevant application area of real-time distributed control promises to result in fundamental and pragmatic results in both areas, and should find immediate applicability to real-world problems.