Learning to Control Autonomous Fleets from Observation via Offline Reinforcement Learning | IEEE Conference Publication | IEEE Xplore