LLM-based Human-like Traffic Simulation for Self-driving Tests

2025-08-23

AI Summary

Existing autonomous driving test platforms exhibit low fidelity in traffic dynamics simulation, particularly failing to capture the diversity and interpretability of human driver behavior; prevailing approaches rely either on hand-crafted rules or narrow-domain data-driven models. Method: This paper proposes HDSim—the first human-like traffic simulation framework integrating cognitive theory with large language models (LLMs). It introduces a hierarchical driver model featuring perception-mediated behavioral regulation, enabling diverse, interpretable representation of driving styles; LLMs guide decision-making indirectly via perception inputs, enhancing behavioral causality. Contribution/Results: Experiments demonstrate that integrating HDSim increases safety-critical failure detection rate by 68% for autonomous driving systems. Moreover, accident attribution becomes more realistic, grounded in coherent causal reasoning with explicit, traceable inference paths.

Technology Category

• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Ensuring realistic traffic dynamics is a prerequisite for simulation platforms to evaluate the reliability of self-driving systems before deployment in the real world. Because most road users are human drivers, reproducing their diverse behaviors within simulators is vital. Existing solutions, however, typically rely on either handcrafted heuristics or narrow data-driven models, which capture only fragments of real driving behaviors and offer limited driving style diversity and interpretability. To address this gap, we introduce HDSim, an HD traffic generation framework that combines cognitive theory with large language model (LLM) assistance to produce scalable and realistic traffic scenarios within simulation platforms. The framework advances the state of the art in two ways: (i) it introduces a hierarchical driver model that represents diverse driving style traits, and (ii) it develops a Perception-Mediated Behavior Influence strategy, where LLMs guide perception to indirectly shape driver actions. Experiments reveal that embedding HDSim into simulation improves detection of safety-critical failures in self-driving systems by up to 68% and yields realism-consistent accident interpretability.

Problem

• Simulating realistic human driver behaviors for autonomous vehicle testing
• Overcoming limitations of existing heuristic and data-driven traffic models
• Enhancing safety failure detection and accident interpretability in simulations

Innovation

• Combining cognitive theory with LLM assistance
• Introducing hierarchical driver model for diverse styles
• Developing Perception-Mediated Behavior Influence strategy