Models and System-Level Coordination Algorithms for Power-in-the-Loop Autonomous Mobility-on-Demand Systems_Alizadeh

Overview: The goal of this project is to devise computational methods for the optimal coordination of autonomous mobility-on-demand (AMoD) systems, with a central focus on accounting for the couplings between the power and transportation networks. AMoD is a rapidly developing mode of transportation wherein self-driving, electric vehicles transport passengers on demand in a given environment (similar to an Uber system, but with self-driving, electric vehicles). Our key observation is that the emerging AMoD technology will give rise to complex couplings between the power and transportation networks over a wide range of temporal and spatial scales (e.g., couplings between charging demand and electricity prices). At the same time, the ability to intelligently route empty vehicles will provide a unique opportunity for joint traffic management and energy dispatch. For example, autonomous electric vehicles could act as distributed mobile storage devices, providing a support mechanism for integration of intermittent renewable energy resources and relieving congestion in energy distribution networks. While power and transportation systems have been extensively studied in isolation, their couplings and joint control remain largely unexplored, particularly in the context of AMoD. Accordingly, this project aims to devise theoretical models and algorithmic tools for the characterization of these couplings and for the data-driven, system-level control of AMoD with the power network in the loop (for short, "power-in-the-loop AMoD systems," or P-AMoD systems).
Target Area: This proposal falls under the target area referred to as "Engineering of Cyber-Physical Systems" in the call. It seeks to provide systematic and reliable analysis and control techniques for next-generation autonomous and electric mobility-on-demand systems. The Core CPS research areas considered by this project are control, autonomy, and real-time systems.
Keywords: Autonomous mobility-on-demand, electric vehicles, transportation system, power system.

Intellectual merit: The intellectual merit of this project is to synergistically integrate and extend algorithmic techniques for vehicle routing, electricity demand management, and multi-agent system coordination in order to generate novel systematic tools for the modeling, analysis, and control of P-AMoD systems. The starting point of this project will be the recent work by the investigators on coordination algorithms for AMoD systems operating in isolation (Dr. Pavone) and on investigating the effects of non-autonomous, electric vehicles on power system demand management (Dr. Alizadeh). Leveraging these results, the key technical idea will be to cast the coupled power and transportation networks in the formal framework of flow optimization, whereby city districts, charging stations, and roads are abstracted as nodes and edges of a graph, and the movements of customers, vehicles, and energy are abstracted as flows over such graph. This project will then devise a control framework that extends techniques from receding horizon control and game-theoretical coordination mechanisms to optimize over the decision variables, e.g., vehicles' routes, charging decisions, and power generation schedules. In particular, this project will generate distributed coordination algorithms and Pigovian taxation schemes to enable independent, possibly noncooperative system operators to (approximately) recover a socially optimal solution without sharing private information. This project will advance the knowledge base in the related fields of vehicle routing, game theory, and multi-agent system theory, which will have broad applicability to several other types of multi-robot systems.
Broader impacts: The broader impact potential of this project stems from its direct application to two of the most important societal infrastructures, namely transportation and energy. The proposed effort will provide currently unavailable tools for the system-wide control of P-AMoD systems and for the analysis of their economic and societal value. Collectively, the results of this project will provide much needed guidelines to transportation stakeholders and policy-makers alike regarding the deployment of autonomous vehicles on a massive scale and their integration with the energy grid. The proposed research plan is also deeply integrated with a number of proposed teaching and training activities. In particular, these activities will impact high school students and freshman undergraduate students by exposing them to and training them on optimization, game theory for robotic applications, and multi-agent systems.