ABSTRACT
This chapter shows how the event-based perturbation analysis calculus may be used in multi-agent systems for determining optimal agent trajectories without any detailed knowledge of environmental randomness. The event-driven paradigm offers an alternative complementary approach to the time-driven paradigm for modeling, sampling, estimation, control, and optimization. The event-driven nature of this approach implies its scalability in the size of an event set, as opposed to the system state space. The event-driven paradigm offers an alternative, complementary look at modeling, control, communication, and optimization. In distributed systems, event-driven mechanisms have the advantage of significantly reducing communication among networked components without affecting desired performance objectives. Event-driven approaches are also attractive in receding horizon control, where it is computationally inefficient to reevaluate a control value over small time increments as opposed to event occurrences defining appropriate planning horizons for the controller. A hybrid system consists of both time-driven and event- driven components.
