Session: Control and Optimization of AVs for Transportation Solutions
Chair: Jeff Ban & Ruolin Li
Cluster: Optimization Applications (Communication, Energy, Health, ML, ...)
Talk 1: Cooperative Optimization of Traffic Signals and Mixed flow of Connected/Automated Vehicles and Human Driven Vehicles
Speaker: Shakiba Naderian
Abstract: Transportation is under rapid transformation with emerging technologies and systems. On the one hand, vehicles are equipped with advanced communication and automation capabilities, leading to connected and automated vehicles (CAVs). On the other hand, infrastructure (such as traffic signals and intersections) is increasingly installed with sensing and data collection systems (such as video cameras, Lidars, edge computing devices, etc.), enabling robust and fast data collection, sharing, and control of traffic flow. Naturally, infrastructure (e.g., traffic signals) and vehicles (e.g., CAVs) should be jointly optimized and controlled in future urban areas to improve safety, mobility, and other related goals. This research concerns about the models and algorithms of cooperative optimization and control of traffic signals and mixed flow of CAV and human driven vehicles (HDVs). The performance of the models and algorithms are tested in simulation, digital twins, and the Mcity 2.0 mixed reality testing environment.
Talk 2: Safety Guaranteed Robust Multi-Agent Reinforcement Learning with Hierarchical Control for Connected and Automated Vehicles
Speaker: Zhili Zhang
Abstract: We address the problem of coordination and control of Connected and Automated Vehicles (CAVs) in the presence of imperfect observations in mixed traffic environment. A commonly used approach is learning-based decision-making, such as reinforcement learning (RL). However, most existing safe RL methods suffer from two limitations: (i) they assume accurate state information, and (ii) safety is generally defined over the expectation of the trajectories. It remains challenging to design optimal coordination between multi-agents while ensuring hard safety constraints under system state uncertainties (e.g., those that arise from noisy sensor measurements, communication, or state estimation methods) at every time step. We propose a safety guaranteed hierarchical coordination and control scheme called Safe-RMM to address the challenge. Specifically, the high-level coordination policy of CAVs in mixed traffic environment is trained by the Robust Multi-Agent Proximal Policy Optimization (RMAPPO) method. Though trained without uncertainty, our method leverages a worst-case Q network to ensure the model's robust performances when state uncertainties are present during testing. The low-level controller is implemented using model predictive control (MPC) with robust Control Barrier Functions (CBFs) to guarantee safety through their forward invariance property. We compare our method with baselines in different road networks in the CARLA simulator. Results show that our method provides best evaluated safety and efficiency in challenging mixed traffic environments with uncertainties.
Talk 3: Game-Theoretic Lane Choice at Highway Weaving Ramps: The Role of AV Altruism
Speaker: Ruolin Li
Abstract: Highway weaving ramps are notorious bottlenecks in modern traffic networks, where merging, exiting, and through flows co-exist in complex, often conflicting ways. In our work, we propose a comprehensive game-theoretic model that predicts the collective lane-changing behavior of mainline vehicles as they approach these high-conflict zones. By modeling drivers’ choices, whether to bypass the merging and exiting chaos by switching lanes or to remain in their current lane, using a concise set of parameters calibrated with minimal traffic data, our approach achieves remarkable predictive accuracy as demonstrated by microscopic SUMO simulations. We further introduce a two-level Stackelberg game framework tailored for mixed traffic weaving ramps that incorporate connected autonomous vehicles (CAVs). At the upper level, we govern the proactive, social-optimal lane-changing strategies of CAVs, while the lower level captures the reactive, self-interested behavior of human-driven vehicles (HDVs). Our analysis quantifies the optimal degree of altruism under varying CAV penetration rates, and reveals the delicate balance between individual benefits, fleet advantages, and societal gains, uncovering the interplay between selfish and altruistic driving behaviors. This adaptable framework offers a powerful tool for diagnosing and alleviating bottlenecks in a series of traffic scenarios such as weaving-ramps. Our findings not only deepen our understanding of AV-HV interactions but also pave the way for smarter, more efficient traffic management strategies in mixed autonomy environments.
