Collaborative Research: CPS: Medium: Robust to Early Termination Optimization for Safe and Reliable Control of High Performance Cyber-Physical Systems

About This Grant

Cyber-physical systems (CPS), for example, autonomous vehicles and delivery drones, consist of computational units tightly integrated with their physical environments. They are often built using small, constrained platforms and multiple control tasks shape the common platform. Thus the computation that is available to each control task is limited and may vary over time, which threatens to compromise the quality of control. This project will address this challenge through co-design of control algorithms and real-time scheduling techniques for control tasks. The new control algorithms developed in this project will be robust to early termination, with guarantees on the quality of control, and the scheduling frameworks will be capable of dynamically adapting scheduling decisions in response to changes in computational demand. The developed techniques will be applied to automated drone delivery in collaboration with industrial partners. The hand-on experimentation plan enables technology transfer to commercial delivery applications, as well as provides a valuable educational tool for engineering students studying robotics and autonomous systems. The approach will target optimization based control algorithms, such as model-predictive control and apply novel solvers based on state-of-the-art Robust to EArly termination oPtimization (REAP). The core idea of REAP is to construct a continuous-time dynamical system whose trajectory converges to the optimal solution, while a sub-optimal and feasible solution is guaranteed even in the event of early termination. Towards achieving this, the project will investigate i) closed-loop stability guarantees and discrete-time implementation; ii) proactive and safe real-time scheduling in CPSs; iii) cooperative computation-aware distributed model predictive control; and iv) control of systems subject to time-varying constraints. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.