FlatPack

Abstract

Efficient use of automated planning and scheduling has been proposed in the Industry 4.0 context involving multiple robotic agents. Temporal planners are typically employed with durative actions included for concurrent plan scheduling and execution. However, the planned durations are either mean values or within closed intervals, which can be easily violated during execution due to uncontrollable events or agent deviations. These variations trigger violation of safety properties in multi-agent scenarios, mandating further verification and control during execution. In this paper, we propose Flat-Pack, a framework to generate flexible temporal plans that can ensure satisfaction of specified properties during execution. Temporal plans are translated to timed-game automata models for verification in Uppaal-Tiga prior to dispatch. Further, using the strategy synthesized by Uppaal-Tiga, controlled plan executions may be performed that are resilient to temporal deviations. The optimal number of controllable transitions are further derived with tradeoffs against control complexity and end-to-end plan execution timelines. These aspects are demonstrated over an Industry 4.0 warehouse use case involving coordination of multiple robots.