Co-design of Platform and Control in NCS Using an Overrun Framework
Abstract:
This research is focused on the co-design of platform and control of Cyber-Physical Systems (CPS). Limited and shared resources among several control and non-control applications in CPS introduce delays in transmitted messages, which in turn can degrade system performance or even cause instabilities. While a worst-case delay based design can accommodate such delays, they often are pessimistic and lead to an overdesign as worst case delays happen infrequently. A priority-based approach can permit varying delays, but this can lead to inefficient use of resources as well. In this research, we propose an overrun framework to carry out a co-design of control and implementation platform that leads to high control performance and optimal resource utilization. We use a two-parameter model for delay thresholds, with the two parameters corresponding to a small and a large delay bound, respectively. A nominal-skip-abort switching control strategy is proposed depending on the relation between the delay experienced by a given message and these two delay values. The skip strategy implies that the next control message is skipped, and the abort one implies that the current control message computed is aborted. The resulting switching closed-loop system is shown to be stable under conditions on the skip-rate and abort-rate. The advantages of such an overrun strategy are not only a guaranteed control performance but also efficient resource utilization as both skip and abort allow the computed input to have enough time to affect the system's performance while freeing resources for lower priority applications. Using such an overrun algorithm and platform analysis, a co-design of control and platform is proposed that optimizes the parameters of this overrun strategy. A case study of automotive applications shows that the co-design results in a 40% reduction in the requisite bandwidth of the implementation platform, while achieving desired control performance.
- PDF document
- 1.08 MB
- 66 downloads
- Download
- PDF version
- Printer-friendly version