Integrated distributed control of cyber-physical systems

Project members:Ir. A. Firooznia (Amir), B. De Schutter (Bart)
Keywords:Cyber-physical systems, Hybrid and nonlinear systems, Distributed control, Scheduling
Sponsored by:STW

In Cyber-Physical Systems (CPS), embedded computers monitor, control and coordinate physical processes via networked communication. As a consequence, CPS are systems with a tight integration of the discrete world of computation, communication and control (the cyber part) with the continuous world of physical processes (the physical part). For example, in the physical world, time and concurrency are intrinsic, whereas in the computing domain time and concurrency are engineering concerns.

Despite the need for an integrated design in high-tech industries uniting the discrete and continuous worlds, the corresponding scientific disciplines have largely developed independently. As a result, assumptions made in one domain often cannot be realized, or only at considerable cost, by the other domain. For example, ideal communication infrastructures without latencies and jitter are often assumptions for the feedback control design while in practice this is impossible to realize in a network or only at prohibitive costs.

A recurring challenge in CPS is to manage a flow of physical or virtual objects (pages, wafers, images, radio frames, audio samples) under strict timing and quality constraints, with limited computational resources and in close interaction with a physical environment that can only be controlled to a limited extent. A model-driven approach to integrally design the physical architecture of the CPS, the scheduling of the flow of objects, and the distributed control to execute the schedules is one way to cope with the complexity and reliability challenges.

In any professional printer, the paper path is one of the key components. It is responsible for getting the substrate (which may be paper, textile, or various other materials) to the right place at the right time, respecting physical constraints with respect to heating of the substrate, cooling and drying the print, speed and acceleration, etc. Defining nominal schedules for various substrate sizes and print trajectories and controlling the execution of the schedule is a challenge. In current industrial practice, paper path design is mostly a manual process supported with simulations. In this project, we pursue the automation of the process, focusing on design-time construction of nominal schedules (possibly per path segment and considering multiple objectives such as performance and power dissipation), run-time adaptive construction of an actual schedule, and synthesis of a distributed control strategy to execute the schedule.

Due to the complexity of the given problem, we will develop computationally efficient control approaches such as decentralized and distributed model predictive control. Moreover, because of the hybrid nature of our problem, methods from mixed logical dynamical systems, hybrid automata, and mixed-integer optimization are used and further extended in this project. Another challenge we will address, is to make the design robust against variations and to cope with varying environmental conditions, wear and tear, failing components, etc.

© Copyright Delft Center for Systems and Control, Delft University of Technology, 2017.