Michael Beetz, and Maren Bennewitz
This paper investigates issues in the integration of planning, scheduling, and plan execution in the context of control systems for autonomous robot office couriers. Such control systems have to meet challenging requirements. They must avoid late deliveries and achieve long-term efficiency instead of restricting themselves to the optimization of problem solving episodes. Becauserobots acting in human working environments are often forced to schedule their activities without having complete information their control systems must reliably and efficiently execute scheduled activity in changing and partly unknown situations. In particular, the control systems must be able to exploit opportunities, flexibly avoid problems, and integrate command revisions for currently active tasks. We propose to implement such control systems for scheduled activity by employing concurrent reactive plans that, while performing their actions, reschedule the course of action whenever necessary. The plans have a modular and transparent representation which permits easy transformations by transformation rules. It is these rules that implement the controller’s scheduling and schedule repair methods.