Investigation of Real-Time Task Scheduling on Robot Fleets with Reconfigurable Actuators

Abstract

Multi-Fleet Scheduling (MFS) is concerned with the issue of assigning tasks to a swarm of mobile robotic agents. In this paper, MFS of tasks using a novel class of mobile agents with reconfigurable modular actuators is proposed and analyzed. MFS is split into two regimes, static and dynamic, where the static regime does not allow real-time reconfiguration of agent actuators. Most pre-existing robotic agents are compatible with the static multi-fleet scheduling (S-MFS) regime, whereas the novel agents being investigated here are capable of using dynamic multi-fleet scheduling (D-MFS). Solutions to both problems are compared, and it is shown that in the worst case scenario, given some set of agents and tasks available at known start times, D-MFS finds the same optimal schedule as S-MFS, whereas D-MFS can be used to find more optimal solutions in some conditions. It is also shown that D-MFS may not always be optimal depending on the arrival of previously unknown a-periodic tasks, as D-MFS provides the optimal schedule for a specific fleet of robots accomplishing a set of tasks for some scheduling algorithm and cost function. By defining and exploring the D-MFS problem, this work paves the way for future investigations in task-prediction, efficient large-scale scheduling algorithms, and novel robot manufacturing capabilities.