Output-Based Control and Scheduling of Resource-Constrained Processes

Modern control systems are usually realized as embedded systems. This typically leads to resource-constrained control systems due to constraints on computing speed, memory size or communication bandwidth. In this thesis, scheduling and control strategies as well as event-triggered control strategies are addressed for handling the limited computation and communication resources. All proposed strategies are designed in a way, such that they can be realized based on output information.

Part I focuses on scheduling and control design methods for distributing efficiently the limited computation and communication resources. First, an integrated control and scheduling design is proposed for controlling several plants over one share communication network, which assumes full state measurement. By including a switched observer the control and scheduling can be realized based on output information. Second, a strategy for scheduling a set of PI control tasks with limited computation resources is developed. The focus lies on the scheduling of PI control tasks with low computational overhead. Both presented strategies are online scheduling algorithms, where the scheduling decision is made based on output-feedback information of the plants. This allows a reactiveness to disturbances and reference changes and, hence, an improved control performance compared with offline scheduling algorithms.

In Part II, a novel event-triggered PI control concept is developed for setpoint tracking problems. The focus lies on reducing the resource utilization, especially the communication over the feedback link, while maintaining a certain level of performance. An event-triggered PI control strategy is proposed, where both the event-triggering condition and the PI controller are realized periodically, such an implementation can be easily realized in a standard time-sliced embedded software architecture. Further, a control synthesis is presented, which allows to design the control parameters taking the event-triggered implementation into account.

All proposed strategies in Part I and Part II are evaluated by simulations and experimental studies in comparison with related approaches from literature, to demonstrate the effectiveness.