Towards Proactive Replanning for Multi-Robot Teams

Abstract

Rather than blindly following a predetermined schedule, human workers often change tasks in order to assist a coworker experiencing difficulties. We are examining how this notion of ``helpful'' behavior can inspire new approaches to online plan execution and repair in multi-robot systems. Specifically, we are investigating proactive replanning, which attempts to predict problems or opportunities and adapt to them by shifting agents between executing tasks. By continuously predicting a task's remaining duration, a proactive replanner is able to accommodate upcoming problems or opportunities before they manifest themselves. One way to do so is by adding or removing agents to or from the various executing tasks, allowing the planner to balance a schedule in response to the realities of execution. We have developed a proof-of-concept system that implements duration prediction and modification of existing tasks, yielding simulated executed makespans as much as 32% shorter than possible without these capabilities.