Cooperative Driving of Connected Autonomous Vehicles in Heterogeneous Mixed Traffic: A Game Theoretic Approach (2305.03563v1)

Abstract: High-density, unsignalized intersection has always been a bottleneck of efficiency and safety. The emergence of Connected Autonomous Vehicles (CAVs) results in a mixed traffic condition, further increasing the complexity of the transportation system. Against this background, this paper aims to study the intricate and heterogeneous interaction of vehicles and conflict resolution at the high-density, mixed, unsignalized intersection. Theoretical insights about the interaction between CAVs and Human-driven Vehicles (HVs) and the cooperation of CAVs are synthesized, based on which a novel cooperative decision-making framework in heterogeneous mixed traffic is proposed. Normalized Cooperative game is concatenated with Level-k game (NCL game) to generate a system optimal solution. Then Lattice planner generates the optimal and collision-free trajectories for CAVs. To reproduce HVs in mixed traffic, interactions from naturalistic human driving data are extracted as prior knowledge. Non-cooperative game and Inverse Reinforcement Learning (IRL) are integrated to mimic the decision making of heterogeneous HVs. Finally, three cases are conducted to verify the performance of the proposed algorithm, including the comparative analysis with different methods, the case study under different Rates of Penetration (ROP) and the interaction analysis with heterogeneous HVs. It is found that the proposed cooperative decision-making framework is beneficial to the driving conflict resolution and the traffic efficiency improvement of the mixed unsignalized intersection. Besides, due to the consideration of driving heterogeneity, better human-machine interaction and cooperation can be realized in this paper.

• The paper introduces a novel cooperative decision-making framework using a Normalized Cooperative Level-k game to allocate right-of-way effectively at unsignalized intersections.
• The paper leverages real-world traffic data and inverse reinforcement learning to simulate heterogeneous human-driven vehicle behaviors, enhancing safety and efficiency.
• The paper validates its framework through extensive simulations that demonstrate improved average speeds, reduced delays, and higher safety metrics compared to traditional methods.

"Cooperative Driving of Connected Autonomous Vehicles in Heterogeneous Mixed Traffic: A Game Theoretic Approach" (2305.03563)

Introduction

The paper investigates cooperative driving of Connected Autonomous Vehicles (CAVs) amidst heterogeneous mixed traffic characterized by high-density unsignalized intersections. These intersections pose challenges in efficiency and safety due to the intricate interactions between CAVs and Human-driven Vehicles (HVs). The authors propose a novel cooperative decision-making framework leveraging game theory to resolve conflicts effectively.

Framework for Cooperative Driving

The proposed framework synthesizes the interaction dynamics between CAVs and HVs using a Normalized Cooperative game combined with a Level-k game (NCL game). Central to this approach is effectively allocating Right of Way (ROW) to maximize system efficiency and safety.

• Level-k Game: Models varying reasoning depths among vehicles, where each vehicle plans its trajectory based on predictions of others' movements at different levels of reasoning.
• Normalized Cooperative Game: Determines the optimal ROW allocation that achieves a system optimal solution, correcting for the unique characteristics of going straight versus turning vehicles, which traditionally receive different priority in unsignalized intersections.

  Figure 1: Thumbnail of the high-density, unsignalized intersection with heterogeneous HVs involved.

Reproduction of Heterogeneous HV Decisions

A fundamental aspect of this study involves simulating HV decisions in mixed traffic using real-world interaction data. The paper employs classification techniques to categorize drivers into distinct behavior profiles—aggressive, normal, and conservative—and mathematically models these using non-cooperative game theory.

• Data Collection: Interaction data was sourced from a real-world intersection study, leveraging Post Encroachment Time (PET) as a key metric to understand HV behavior.
• Behavior Modeling: Inverse Reinforcement Learning (IRL) was used to calibrate the decision-making preferences, optimizing reward functions that represent efficiency, comfort, and safety priorities.

Algorithmic Implementation

The authors detail a multi-layer hierarchical approach to CAV decision-making and trajectory planning, integrating Level-k game logic into trajectory generation via Lattice planner.

Figure 2: Cooperative driving framework in mixed traffic.

• Lattice Planner: Utilizes transformed coordinates and sampling strategies to generate collision-free trajectories. The optimization process maximizes the reward functions calibrated earlier through IRL, ensuring CAVs interact effectively with HVs.

Simulation Results and Validation

Extensive simulations validate the framework's effectiveness under varied conditions including different CAV penetration rates (ROP) and heterogeneous driver compositions. Key performance metrics include average travel speed, total delay, and PET, demonstrating significant efficiency and safety improvements.

Figure 3: Average travel speed comparison of control method under different lane volume.

• Comparison with Traditional Methods: The NCL game outperformed reservation-based control methods such as FCFS and Batch-strategy, especially in high-density scenarios, showing resilience and improved throughput and delay metrics.
• Safety Analysis: PET analysis revealed a higher concentration and elevated minimum PET in scenarios with increased CAV penetration, indicating better conflict resolution and overall traffic safety.

Conclusion

The game-theoretic framework proposed in this paper offers a robust approach for managing CAVs in complex, unsignalized intersection environments. It advances current methodologies by introducing cooperative elements that account for heterogeneous traffic dynamics. Future work could explore scaling this approach to larger, more varied traffic networks and integrating real-time data from infrastructure sensors for dynamic adaptability.