An Introduction to Event-Based Optimization: Theory and Applications

In this chapter, we review the basic idea of event-based optimization (EBO), which is specifically suitable for policy optimization of discrete event dynamic system (DEDS). With decisions based on certain events instead of on the system state, the number of potentials to be estimated is usually much smaller than the size of the state space. Performance difference and derivative formulas are developed for event-based policies. Under certain assumptions, policy iteration algorithms for EBO can be developed and they may converge to a globally optimal event-based policy. However, gradientbased optimization algorithms can always be applied and they usually converge to a local optimum in the parameterized policy space. We illustrate the EBO method by a material handling (MH) problem. We hope this chapter may bring insights for the study of event-based control, decision-making, and optimization in more general situations.