SAMoD: Shared Autonomous Mobility-on-Demand using Decentralized Reinforcement Learning

Abstract

Shared mobility-on-demand systems can improve the efficiency of urban mobility through reduced vehicle ownership and parking demand. However, some issues in their implementations remain open, most notably the issue of rebalancing non-occupied vehicles to meet geographically uneven demand, as is, for example, the case during the rush hour. This is somewhat alleviated by the prospect of autonomous mobility-on-demand systems, where autonomous vehicles can relocate themselves; however, the proposed relocation strategies are still centralized and assume all vehicles are a part of the same fleet. Furthermore, ride-sharing is not considered, which also has an impact on rebalancing, as already occupied vehicles can also potentially be available to serve new requests simultaneously. In this paper we propose a reinforcement learning-based decentralized approach to vehicle relocation as well as ride request assignment in shared mobility-on-demand systems. Each vehicle autonomously learns its behaviour, which includes both rebalancing and selecting which requests to serve, based on its local current and observed historical demand. We evaluate the approach using data on taxi use in New York City, first serving a single request by a vehicle at a time, and then introduce ride-sharing to evaluate its impact on the learnt rebalancing and assignment behaviour.

Shared Autonomous Mobility on Demand: A Learning-Based Approach and Its Performance in the Presence of Traffic Congestion

“…To address this gap, we evaluate the influence of regular traffic, i.e., private cars, on the behaviour of a fleet of SAVs, as well as the influence of varying sizes of SAV fleets on regular traffic. We extend our previous work SAMoD [5] in which we proposed a fully decentralized RL approach to ride-sharing and vehicle rebalancing. We investigate whether SAMoD's achieved behaviour holds in the presence of congestion, and what effect does it have on the overall traffic flow.…”

“…In our previous work [5], we introduced a decentralized Shared Autonomous Mobility-on-Demand system (SAMoD), a reinforcement learning (RL) based decentralized approach to vehicle rebalancing as well as ride request assignment in shared mobility-on-demand systems. In this paper we reiterate the main features of this approach, but go further by extending SAMoD algorithm to make request assignment and rebalancing decisions based on real-world road network and in the presence of congestion simulated in micro-simulator SUMO.…”

“…The only approaches that consider congestion with ridesharing and/or rebalancing are [9], [12], [13], [14]. However [15] Hörl et al [13] Zhang et al [6] Simonetto et al [7] Lu et al [8] Levin et al [9] Martinez et al [16] Fagnant and Kockelman [3] Fiedler [10] Alonso-Mora et al [4] Wen et al [11] Vosooghi et al [12] Ruch et al [14] SAMoD Guériau and Dusparic [5] This paper none uses a congestion-aware routing strategy to react to observed congestion level. [9], [12] only account for congestion generated by the fleet of SAVs themselves rather than other vehicles, and in [14], [13] the agent-based simulation used relies on a simplified queue-based model for traffic congestion instead of more realistic car-following behaviours.…”

Literature Review Toward Decentralized Railway Traffic Management

“…A solid approach for shared mobility using a reinforcement learning algorithm is proposed in [38]: vehicles are modelled as autonomous agents, giving advices when a driver is in control and directing otherwise, learning how to behave by using rewards. The paper seeks to tackle two main problems: assigning vehicles to users and performing vehicle rebalance, i.e., redistributing idle vehicles to high demand areas.…”

Multiagent Reinforcement Learning-Based Taxi Predispatching Model to Balance Taxi Supply and Demand

“…ere are also some researchers who use reinforcement learning to achieve their goals. Guériau and Dusparic [32] propose a reinforcement learning-based decentralized approach to vehicle relocation as well as ride request assignment in shared mobility-on-demand systems. Each vehicle autonomously learns its behaviour, including both rebalancing and selecting which requests to serve, based on its local current and observed historical demand.…”