Data-driven intermittent control of linear systems

This paper is concerned with the data-driven intermittent control problem for unknown linear systems. By intermittent control, one refers to a controller that is activated during certain time intervals and deactivated otherwise. Based on the triggering methods of activation and deactivation, it can be categorised into time-based and event-based types. First, a data-driven time-based intermittent control (TbIC) framework is proposed for the scenario where the controller's activation and deactivation time sequences are aperiodic. Under this framework, the control gain can be directly solved from the noisy dataset. Subsequently, to eliminate unnecessary control intervals, an event-triggered mechanism from data is proposed to autonomously activate and deactivate the controller, based on which event-based intermittent control (EbIC) is further developed. Furthermore, TbIC and EbIC strategies with zero-order-hold updates are designed to reduce the number of controller updates during control intervals. Numerical examples demonstrate that, compared with the system-identification-based method, the proposed data-driven intermittent control is superior in terms of stability and control cost savings.