Decision-driven scheduling

This paper presents a scheduling model, called decision-driven scheduling, elaborates key optimality results for a fundamental scheduling model, and evaluates new heuristics solving more general versions of the problem. In the context of applications that need control and actuation, the traditional execution model has often been either time-driven or event-driven. In time-driven applications, sensors are sampled periodically, leading to the classical periodic task model. In event-driven applications, sensors are sampled when an event of interest occurs, such as motion-activated cameras, leading to an event-driven task activation model. In contrast, in decision-driven applications, sensors are sampled when a particular decision must be made. We offer a justification for why decision-driven scheduling might be of increasing interest to Internet-of-things applications, and explain why it leads to interesting new scheduling problems (unlike time-driven and event-driven scheduling), including the problems addressed in this paper.