Enhancing Social Decision-Making of Autonomous Vehicles: A Mixed-Strategy Game Approach With Interaction Orientation Identification

TL;DR

The paper tackles improving social decision-making for autonomous vehicles in mixed traffic, particularly at unsignalized intersections. It introduces a three-module framework—Interaction Orientation Identification, Mixed-Strategy Game Modeling, and Expert Mode Learning—that quantifies social tendencies and solves non-cooperative games with future-state payoffs, while learning from real expert driving data. The Interaction Orientation module computes an Interaction Transient State Index and IO score from environmental and trajectory cues; the payoff design integrates safety, efficiency, and stochastic disturbances; the expert-mode library enables adaptive decisions via Lemke-Howson solutions. Validation on Argoverse2 and SinD data, plus human-in-the-loop experiments, demonstrates improved decision timing, accuracy, efficiency, and safety. These results suggest the approach can enhance AV-HV interaction at complex intersections and generalizes to more dynamic social driving contexts.

Abstract

The integration of Autonomous Vehicles (AVs) into existing human-driven traffic systems poses considerable challenges, especially within environments where human and machine interactions are frequent and complex, such as at unsignalized intersections. To deal with these challenges, we introduce a novel framework predicated on dynamic and socially-aware decision-making game theory to augment the social decision-making prowess of AVs in mixed driving environments. This comprehensive framework is delineated into three primary modules: Interaction Orientation Identification, Mixed-Strategy Game Modeling, and Expert Mode Learning. We introduce 'Interaction Orientation' as a metric to evaluate the social decision-making tendencies of various agents, incorporating both environmental factors and trajectory characteristics. The mixed-strategy game model developed as part of this framework considers the evolution of future traffic scenarios and includes a utility function that balances safety, operational efficiency, and the unpredictability of environmental conditions. To adapt to real-world driving complexities, our framework utilizes a dynamic optimization framework for assimilating and learning from expert human driving strategies. These strategies are compiled into a comprehensive strategy library, serving as a reference for future decision-making processes. The proposed approach is validated through extensive driving datasets and human-in-loop driving experiments, and the results demonstrate marked enhancements in decision timing and precision.

Figures (13)

• Figure 1: The interaction scenario.
• Figure 2: The social decision-making game framework.
• Figure 3: Schematic Diagram of Trajectories for Interacting Parties: (a) Normal Driving, (b) Lateral Deviation in Left-Turning Vehicle, (c) Lateral Deviation in Straight-Going Vehicle.
• Figure 4: Space-Time (ST) Diagram Centered on the Straight-Going Vehicle: (a) ST Trajectories and Interaction Partner's Occupied Area under Normal Conditions, (b) ST Diagram following Lateral Deviation of the Left-Turning Vehicle, (c) Determining Acceleration Boundaries based on the ST Diagram.
• Figure 5: The relationship between IO and ITSI and $S_{norm}$