Congestion Management for Mobility-on-Demand Schemes that use Electric Vehicles

TL;DR

This paper introduces congestion management algorithms for Electric Vehicles in a Mobility-on-Demand system, optimizing trip assignments to balance demand and supply using both optimal and heuristic approaches.

Contribution

It proposes a novel mixed-integer programming solution and real-time heuristic algorithms for EV trip assignment in MoD schemes, addressing management challenges.

Findings

Heuristic (a) improves trip execution by up to 4.8% over (b) and 11.5% over (c).

The online algorithms perform close to the optimal offline solution.

The approach scales to systems with dozens of EVs and hundreds of requests.

Abstract

To date the majority of commuters use their privately owned vehicle that uses an internal combustion engine. This transportation model suffers from low vehicle utilization and causes environmental pollution. This paper studies the use of Electric Vehicles (EVs) operating in a Mobility-on-Demand (MoD) scheme and tackles the related management challenges. We assume a number of customers acting as cooperative agents requesting a set of alternative trips and EVs distributed across a number of pick-up and drop-off stations. In this setting, we propose congestion management algorithms which take as input the trip requests and calculate the EV-to-customer assignment aiming to maximize trip execution by keeping the system balanced in terms of matching demand and supply. We propose a Mixed-Integer-Programming (MIP) optimal offline solution which assumes full knowledge of customer demand and an equivalent online greedy algorithm that can operate in real time. The online algorithm uses three alternative heuristic functions in deciding whether to execute a customer request: (a) The sum of squares of all EVs in all stations, (b) the percentage of trips' destination location fullness and (c) a random choice of trip execution. Through a detailed evaluation, we observe that (a) provides an increase of up to 4.8% compared to (b) and up to 11.5% compared to (c) in terms of average trip execution, while all of them achieve close to the optimal performance. At the same time, the optimal scales up to settings consisting of tenths of EVs and a few hundreds of customer requests.