Logo: to the web site of Uppsala University

Wireless control systems have gained considerable attention in recent years due to their numerous advantages, including increased flexibility and scalability, reduced wiring complexity, and cost-efficiency. Despite these benefits, the use of communication networks in control loops poses various challenges, such as sampled data, latency, packet dropouts, etc. Moreover, in contrast to wired connections, wireless control systems present a new essential challenge – energy consumption. To mitigate this issue, we can use energy harvesting, which involves the conversion of ambient energy sources into usable electrical energy. However, the successful use of harvesting technologies in wireless control systems requires appropriate control methods, transmission policies, accurate harvesting models, and a good understanding of communication channel behavior.

In this thesis, we present two approaches for designing suitable transmission policies. The first one is event-triggered control, which can reduce the number of transmissions over the network and guarantee system stability. We use the input delay approach and the Lyapunov technique to obtain exponential stability conditions formulated as linear matrix inequalities. If the system contains external inputs, we analyze it in terms of input-to-state stability. Additionally, we investigate event-triggered control with packet dropouts. For this case, we represent the initial system as a stochastic impulsive system and obtain conditions that guarantee mean-square exponential stability.

The second approach is based on reinforcement learning, where the transmission policy is designed as a trade-off between control performance and energy consumption. Implementing such policies requires knowledge of harvested and transmitted energy. The latter, in turn, is highly dependent on the communication channel state, which therefore needs to be estimated as well. These two issues are also addressed in this thesis.