Infrastructure Design for Electric and Autonomous Vehicles

Abstract

This thesis focuses on infrastructure design for the disruptive transportation technologies of electric vehicles (EVs) and autonomous vehicles (AVs) to enable their adoption at large scale. Particularly, two EV-related problem frameworks concerning the spatial distribution of charging stations and their respective capacity levels are studied, and a new problem is introduced to determine the optimal deployment of AV lanes and staging facilities to enable shared autonomous transportation in urban areas.

The first problem is centered around determining optimal locations of fast-charging stations to enable long-distance transportation with EVs. A new mathematical model is developed to address this problem. This model not only determines optimal facility locations but also finds optimal routes for every origin-destination (OD) trip which follows the path that leads to the minimum total en route recharging. Through computational experiments, this model is shown to outperform the widely used maximum and set cover problem settings in the literature in terms of several routing-related performance measures. A Benders decomposition algorithm is developed to solve large-scale instances of the problem. Within this algorithm, a novel subproblem solution methodology is developed to accelerate the performance of the classical Benders implementation. Computational experiments on real-world transportation networks demonstrate the value of this methodology as it turns out to speed the classical Benders up to 900 times and allows solving instances with up to 1397 nodes.

The second problem extends the previous one by seeking to determine EV charging station locations and capacities under stochastic vehicle flows and charging times. It also considers the route choice behavior of EV users by means of a bilevel optimization model. This model incorporates a probabilistic service requirement on the waiting time to charge, and it is studied under a framework where charging stations operate as M/M/c queuing systems. A decomposition-based solution methodology, that uses a logic-based Benders algorithm for the location-only problem, is developed to solve the proposed bilevel model. This methodology is designed to be versatile enough to be tailored for the cooperative or uncooperative EV user behavior. Computational experiments are conducted on real-life highway networks to evaluate how service level requirements, deviation tolerance levels, and route choice behavior affect the location and sizing decisions of charging stations.

The third problem entails the staging facility location and AV lane deployment problem for shared autonomous transportation. The proposed problem aims to find the optimal locations of staging facilities utilizing a bi-objective model that minimizes total travel distance and the total AV travel not occurring on AV lanes with respect to a given AV lane deployment budget and a number of staging facilities to locate. A Benders decomposition algorithm with Pareto-optimal cuts is developed and the trade-offs with optimal solutions on benchmark instances are evaluated. Computational experiments are performed to analyze the effects of AV lane budget, staging facility count, and the objective preferences of decision makers on optimal solutions.