[Paper Review] Human-like Driving Decision at Unsignalized Intersections Based on Game Theory

Daofei Li, Guanming Liu+1 more|arXiv (Cornell University)|Dec 12, 2021

Autonomous Vehicle Technology and Safety4 citations

TL;DR

This paper proposes a game-theoretic framework for human-like driving decisions at unsignalized intersections, using Prospect Theory to map collision risk to subjective payoffs and modeling driver acceleration tendencies probabilistically. The approach achieves 98% success rate in safe two-vehicle interactions while maintaining speed efficiency, validated in both two- and four-vehicle scenarios.

ABSTRACT

Unsignalized intersection driving is challenging for automated vehicles. For safe and efficient performances, the diverse and dynamic behaviors of interacting vehicles should be considered. Based on a game-theoretic framework, a human-like payoff design methodology is proposed for the automated decision at unsignalized intersections. Prospect Theory is introduced to map the objective collision risk to the subjective driver payoffs, and the driving style can be quantified as a tradeoff between safety and speed. To account for the dynamics of interaction, a probabilistic model is further introduced to describe the acceleration tendency of drivers. Simulation results show that the proposed decision algorithm can describe the dynamic process of two-vehicle interaction in limit cases. Statistics of uniformly-sampled cases simulation indicate that the success rate of safe interaction reaches 98%, while the speed efficiency can also be guaranteed. The proposed approach is further applied and validated in four-vehicle interaction scenarios at a four-arm intersection.

Motivation & Objective

• To address the challenge of safe and efficient automated driving at unsignalized intersections with dynamic, unpredictable vehicle interactions.
• To model human-like driving behavior by quantifying the trade-off between safety and speed using subjective payoffs.
• To incorporate dynamic interaction effects through a probabilistic model of driver acceleration tendencies.
• To validate the decision-making framework in complex four-vehicle intersection scenarios.
• To achieve high safety and efficiency in automated driving decisions without traffic signals.

Proposed method

• A game-theoretic decision framework is designed to model interactions between automated vehicles and human-driven vehicles at unsignalized intersections.
• Prospect Theory is applied to transform objective collision risk into subjective driver payoffs, capturing risk perception and behavioral trade-offs.
• Driving style is quantified as a balance between safety (low risk) and speed (high efficiency) through the payoff function.
• A probabilistic model is introduced to represent the stochastic acceleration behavior of human drivers during interactions.
• The payoff function integrates risk perception and behavioral dynamics to guide autonomous vehicle decisions in real-time.
• The framework is evaluated in both two-vehicle limit cases and four-vehicle four-arm intersection scenarios using simulation.

Experimental results

Research questions

• RQ1 How can automated vehicles make human-like decisions at unsignalized intersections with minimal infrastructure?
• RQ2 How can subjective driver payoffs be modeled to reflect the real-world trade-off between safety and speed?
• RQ3 What role does probabilistic modeling of driver acceleration play in improving interaction predictability?
• RQ4 How does the proposed game-theoretic approach perform in complex, multi-vehicle intersection scenarios?
• RQ5 Can the method achieve both high safety and acceptable speed efficiency in dynamic traffic interactions?

Key findings

• The proposed method achieves a 98% success rate in safe two-vehicle interactions across uniformly-sampled simulation cases.
• The framework maintains high speed efficiency while ensuring safety through risk-aware payoff modeling.
• Simulation results demonstrate stable and human-like behavior in limit cases of two-vehicle interaction.
• The approach successfully generalizes to four-vehicle interactions at a four-arm intersection, maintaining robustness.
• The integration of Prospect Theory enables accurate representation of driver risk perception in decision-making.
• The probabilistic model of driver acceleration tendencies enhances the realism and adaptability of the decision algorithm.