X5: Dynamics Centered Optimal and Learning Based Mechanism Transition for Cyber-physical Systems

Operating dynamical systems over communication networks provides enormous opportunities. However, how to utilize these advances for safe, high-performance cyber-physical systems remains widely open. Current systems must cope with the quality of service provided by the communication mechanisms, limiting applications to non-safety-critical, slow tasks. These limitations could, in principle, be overcome by a tight and tailored combined optimization and design of the cyber-physical and the communication systems, which is, however, computational and conceptional currently not possible. To break complexity, we propose that cyber-physical systems explicitly take the available communication mechanisms and predictions of their properties into account for their decisions. To satisfy the performance requirements, the cyber-physical systems should dynamically adapt their behavior, taking the available resources and quality of the communication mechanisms into account. While the methods developed should be general in nature, we focus on problems where the properties of the communication mechanism may explicitly depend on the state of the cyberphysical system itself. Exploiting those degrees of freedom, utilizing predictions of the closed-loop system dynamics and forecasts of the communication mechanism behavior in an optimization-based framework sets the stage and objective for the current proposal's objective. Develop a modular solution, where the decision-maker of the cyberphysical system takes the available communication mechanism and dynamic predictions of their properties into account. The methods should allow leverage of predictions of the cyber-physical systems communication demand to the network, exploit available communication mechanism state prediction, and learn from their actions while providing deterministic or stochastic performance, safety, and stability guarantees.