Reactive Scheduling for Human-Robot Collaboration: An Event-Driven Approach with Adaptive Skill Execution

Abstract

Human-Robot Collaboration opens up new opportunities for small and medium-sized enterprises to meet the demands of mass customization while increasing productivity. However, these companies often lack robotics experts, so they need programming approaches that non-experts can understand. To address this, we previously introduced the Robot Collaboration Language (RCL), a skill-based visual Domain-Specific Language that models workflows as Gantt charts. The execution of such a Gantt-modeled workflow, however, comes with its own challenges. As human timing fluctuates during execution, the temporal constraints between tasks modeled in RCL can only be honored if the schedule is continuously updated and propagated as work progresses. To address these challenges, an event-driven architecture is presented that combines three core mechanisms: a decomposition algorithm based on strongly connected components to detect cyclic temporal dependencies; a linear programming adaptive scheduling to enable feasible time adjustments while accounting for ongoing tasks; and a reactive execution strategy that propagates updates throughout execution. This approach enables a seamless synchronization between human and robot, providing an execution framework that remains robust under real-world temporal uncertainty.