Optimal Real-Time Human Attention Allocation and Scheduling in a Multi-human and Multi-robot Collaborative System

Our earlier work developed a contention-resolving model predictive control (or MPC) framework to optimally schedule a single human operator’s attention to collaborate with multiple robots. In this paper, we present a generalized design that can allocate and schedule the limited human attention in a multi-human and multi-robot collaborative system. We establish a new analytical timing model for multi-resource and multi-task real-time systems, where collaboration tasks can be either preemptive or non-preemptive. Then we derive the condition to predict the moments when contentions occur among the collaboration requests to humans. The contention-resolving MPC is triggered at contention moments to dynamically schedule human attention and determine the robot with which each human operator should collaborate. The optimal schedule, which aims to maximize the overall robots performance, is computed using an event-triggered and sampling-based approach with a weighted decision tree. To further improve computation efficiency, we present a new rule to merge leaves and simplify the decision tree. This paper also introduces a reset time for contention-resolving MPC, shortening the time horizon to search the optimal solution and reduce computational requirements. The effectiveness of the proposed method is verified through simulations.