Co-Design of Multimodal CPS Architectures and Adaptive Controllers

Abstract:

The project is an on-going collaborative effort between Massachusetts Institute of Technology and University of Pennsylvania since 2011. The ultimate goal of the project is to develop techniques for the co-design of multi-modal distributed embedded architectures and adaptive controllers so as to (1) achieve a high quality of control (QoC) in a physical application where multiple physical plants have to be simultaneously controlled in real-time; (2) guarantee a given quality of service (QoS) requirement of other coexisting real-time, non-control applications; and (3) optimize the architectural resource use. Our specific objectives are (i) the development of analysis tools for a multi-modal architecture with verifiable implementation that can guarantee specified QoC and specified QoS requirements subjected to given architectural resource constraints, (ii) the development of synthesis tools for the multi-modal architecture consisting of co-designed modes and mode-transitions together with control strategies, and (iii) validate the multi-modal co-design architecture in a practical setting. To date, we have developed techniques to construct switching controllers that ensure stability of a control system subject to uncertain propagation delays between system components, subject to an assumption about the maximum number of delayed messages within any window of a given size. To establish a co-design framework, we complement these control design techniques with platform analysis and configuration techniques that provide guarantees that the assumption about message delays is satisfied. A case study from the automotive domain demonstrates that the co-design approach enables more efficient use of platform resources compared to a conventional design.