An ITE Seminar with Dr. Ruolin Li: The Potential of Autonomous Vehicles in Future Transportation Systems

On Thursday, November 14th, Dr. Ruolin Li from the University of Southern California's Department of Civil and Environmental Engineering presented a seminar on “The potential of autonomous vehicles in future transportation systems.” Dr. Li, who recently joined USC, shared insights from her ongoing research and discussed collaboration opportunities in this emerging field.

Dr. Li's background includes postdoctoral work at Stanford's Department of Aeronautics and Astronautics and a PhD in mechanical engineering from UC Berkeley. Her research lies at the intersection of game theory, multi-agent systems, and control optimization, aiming to enhance the societal benefits of intelligent transportation systems.

At the forefront of innovations in transportation, Dr. Li's work focuses on modeling the interactions between human drivers and autonomous vehicles, particularly in scenarios like highway on-ramps. She highlighted the challenges posed by selfish human driver behavior and proposed solutions using altruistic autonomous vehicles.

Dr. Li presented her "Key Research Focuses":

1. Modeling human-autonomous vehicle interactions

2. Designing altruistic behavior for autonomous vehicles

3. Optimizing transportation systems under uncertainty

4. Integrating autonomous vehicles into existing infrastructure

She emphasized the importance of considering human factors when designing autonomous vehicle systems. She introduced the concept of an "altruism level" for autonomous vehicles, which balances self-interest with societal benefits. Her research suggests that a certain threshold of altruistic vehicles is necessary to improve overall traffic conditions.

Dr. Li highlighted specific examples of selfish human driver behavior that can lead to suboptimal traffic conditions. One such example she demonstrated in her video simulation was vehicles suddenly bypassing to the left lane. These drivers abruptly change lanes, often at the last moment before a merge or exit, to avoid merging traffic. This behavior can disrupt the flow of both lanes and potentially cause safety issues. This type of driving is problematic because:

1. It can create sudden slowdowns in the left lane as other drivers react to the unexpected lane change.

2. It reduces the efficiency of the merging process by leaving gaps in the right lane that could have been used by merging vehicles.

3. It increases the risk of accidents due to the sudden and often aggressive nature of the maneuver.

Dr. Li used this example to illustrate how individual driver decisions, while potentially beneficial for that specific driver in the short term, can negatively impact overall traffic flow and safety. This type of behavior is one of the key challenges that her research aims to address through the strategic use of autonomous vehicles programmed with more altruistic behavior patterns.

Dr. Li presented a mathematical model to predict driver choices between staying in their lane or bypassing merging traffic. This model uses game theory and Nash equilibrium concepts to characterize the lane-choice equilibrium. She emphasized the importance of developing delay models that are both expressive and easy to calibrate.

One of the key insights from Dr. Li's research is the concept of "altruistic autonomous vehicles." These vehicles are programmed to consider not only their own travel time but also their impact on overall traffic flow. Dr. Li introduced an "altruism level" that ranges from 0 (completely selfish) to 1 (completely altruistic).

Dr. Li presented theoretical thresholds for the proportion of altruistic vehicles needed to improve traffic conditions. She explained that to optimize social traffic conditions, a certain proportion of vehicles (δopt) need to be altruistic, while a smaller proportion is needed just to improve conditions.

An important aspect of Dr. Li's research is the consideration of imperfect measurements in real-world scenarios. She introduced the concept of the "price of anarchy" to quantify the impact of uncertainty on traffic systems. Her findings suggest that pure altruism may not be ideal when measurements are imperfect, and an optimal altruism level can be calculated for different scenarios.

Dr. Li briefly mentioned her ongoing research on toll lane policies for mixed autonomy systems. This project aims to optimize interactive mixed autonomy to design tolling policies, potentially adapting high-occupancy vehicle infrastructure for autonomous vehicles.

During the Q&A session, Dr. Li addressed questions about the practical implementation of altruistic autonomous vehicles and the challenges of imperfect measurements in real-world scenarios. She also discussed the potential for adapting existing high-occupancy vehicle infrastructure to integrate autonomous vehicles seamlessly.

Dr. Li concluded by introducing her newly established Societally Aware Interactive Autonomous Systems (SIAS) lab at USC and inviting collaboration opportunities. She also announced a new graduate-level course, "Game Theory for Interactive Autonomy," to be offered in the upcoming spring semester. This course will cover game theoretic modeling of strategic interactions between autonomous agents and humans, with applications in autonomous driving, collaborative robotics, and multi-agent decision making.

Ending with an inspiring message, Dr. Li emphasized the potential of autonomous vehicles to revolutionize transportation systems and the importance of interdisciplinary collaboration in advancing this field. She encouraged attendees to explore the intersection of game theory and reinforcement learning, which she considers a hot topic in the current field.

USC ITE thanks Dr. Li for sharing her cutting-edge research and valuable insights with aspiring transportation engineers and researchers. Her work promises to shape the future of intelligent transportation systems and offers exciting opportunities for collaboration and further study.