Learning to influence vehicles’ routing in mixed-autonomy networks

Ma, Xiaoyu

Permalink

https://hdl.handle.net/2142/120118 Copy

Description

Title

Learning to influence vehicles’ routing in mixed-autonomy networks

Author(s)

Ma, Xiaoyu

Issue Date

2023-05-03

Director of Research (if dissertation) or Advisor (if thesis)

Mehr, Negar

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Intelligent Transportation Systems
• Automation Technologies for Smart Cities

Abstract

Road networks will soon be shared between human-driven and autonomous cars, i.e., they will operate under mixed vehicle autonomy. In this thesis, we consider control strategies that maximize throughput in mixed-autonomy traffic networks. In particular, we consider two control strategies that can influence vehicles' routing in mixed-autonomy networks such that the overall network performance is improved. First, we propose that unlike human-driven cars, which can be difficult to control, we can assume a level of control over autonomous cars, which provides us with an additional control input to affect traffic networks. We propose that in mixed-autonomy networks, the headway of autonomous cars can be assigned dynamically to influence vehicles' routing and reduce congestion. We argue that in mixed-autonomy networks, the headway of autonomous cars --- and consequently the capacity of link segments --- is not just a fixed design choice; but rather, it can be leveraged as an {infrastructure control} strategy to {dynamically} regulate capacities. Second, we consider a related but different control approach. We investigate the potential of using not constant but time-varying prices or tolls for vehicles to affect the route choices of autonomous or human-driven vehicles. We demonstrate how dynamically setting prices or tolls for traversing certain network links can increase the overall throughput of the network. To achieve these, we model the dynamics of mixed-autonomy traffic networks while accounting for the vehicles' route choice dynamics. We train an RL policy that learns to regulate either the headway of autonomous cars or the price assigned to each link, such that the total travel time in the network is minimized. We will show empirically that our trained policy can not only prevent the inefficiencies that result from selfish route choices but also decrease total travel time.

Graduation Semester

2023-05

Type of Resource

Thesis

Copyright and License Information

Copyright 2023 Xiaoyu Ma

Owning Collections