MIXED AUTONOMy in Transportation: Emerging Challenges and Opportunities

Overview

Although connected and autonomous vehicle (CAV) technologies have enormous potential to enhance network mobility and sustainability in modern traffic networks, their deployment on roads shared with human-driven vehicles requires careful consideration to avoid unintended outcomes in the resulting mixed-autonomy system. For instance, CAV fleets deployed in isolation can increase network mobility threefold by forming platoons with short headways, maintained by synchronous velocity and acceleration control [1]. However, recent work demonstrates that increasing the fraction of autonomous vehicles in a mixed-autonomy network can actually exacerbate congestion when all vehicles are routed selfishly [2]. As another example, although CAVs on a multi-lane highway can execute lane-switching behaviors to stabilize traffic, recent work indicates that the frequency of the lane switches must be carefully designed to avoid tipping the system back into instability. These examples illustrate that carelessly devised strategies for operating CAVs in mixed autonomy can negate the CAVs’ potential to positively impact modern transportation systems. The focus of this workshop is therefore: “How should mixed-autonomy traffic systems be controlled to maximize the benefits of deploying connected and autonomous vehicles?”

Our workshop aims to study the above question by considering various control-theoretic and learning-centric perspectives on the opportunities and challenges posed by large-scale mixed-autonomy traffic systems. Specifically, we aim to bring together speakers representing expertise in systems theory, game theory, and machine learning. We will hear from Dr. Ruolin Li (USC CEE), Dr. Ramtin Pedarsani (UCSB ECE), Dr. Yue Yu (UMN Aerospace), Dr. Roy Dong (UIUC ISE), and Dr. Karen Leung (UW AeroAstro). Together with the contributed posters, these diverse viewpoints will illustrate the broad impacts of strategic thinking in multi-agent environments, and how progress in these settings is transforming the state-of-the-art across the research community in controlling mixed autonomy traffic and beyond.

More information regarding our invited talks can be found at the Invited Presentations tab.

We have divided our workshop into three session types, each of which allows for a different mixture of voices to be heard:

• Invited Talks - These sessions will allow attendees to learn about recent results and current trends from experts in the research areas of routing and pricing ride-sharing services in mixed autonomy networks.

• Interactive Poster Session - This session will showcase the work of young researchers and will provide opportunities for informal discussions between participants and speakers.

• Panel Discussion - This moderated discussion will include our invited speakers. The moderator will have several prepared questions for each panelist. Attendees will be asked to submit questions via a live online discussion board, which the moderator will curate for the panelists. Tentative discussion topics are as follows:

• What are the most pressing current and future challenges in routing mixed-autonomy traffic, and what are the most promising approaches we have to deal with them?

• How do the mathematical foundations of game theory impact the way we approach challenges and opportunities in controlling mixed-autonomy traffic systems?

• What is the biggest opportunity we have in the mixed-autonomy community to have a positive impact in the age of ChatGPT?

• What are the additional challenges that arise in mixed autonomy due to strategic multi-agent interactions between human-driven and autonomous vehicles?

• How does our current understanding of strategic multi-agent decision-making inform the design of environments shared by human-driven and autonomous vehicles?

Our workshop on the ACC 2025 website: https://acc2025.a2c2.org/program/workshops#session-11-10

Areas of interest: Routing, Tolling, Stability Analysis, and Equilibrium Design for Mixed-Autonomy Traffic Systems

Techniques include: Hybrid Systems and Control, Game theory, Mechanism Design, Optimization, Probability and Statistics, Bayesian inference, Robotics, Reinforcement learning, Generative Models etc.