A Human-Like Pedestrian Model for Automated Driving Simulations

2026-09-24

AI Summary

This study addresses the limitations of existing pedestrian models in adapting to human-like behaviors within complex traffic environments and their difficulty in covering safety-critical scenarios. We propose a theory-constrained Partially Observable Markov Decision Process (POMDP) interaction framework that integrates deep reinforcement learning with domain randomization to model human perceptual, cognitive, and motor decision-making mechanisms. This approach enables high-fidelity reproduction of anthropomorphic behaviors in complex settings such as multi-lane roads. The proposed model successfully replicates a wide range of empirical findings, and the learned policies demonstrate strong cross-environment transferability. Ultimately, this work provides reliable support for the development and safety evaluation of autonomous driving systems.

Abstract

Automated vehicles must be able to interact with pedestrians safely and efficiently across diverse traffic situations. Although driving simulators offer a scalable testbed for learning such capabilities, existing theory-inspired pedestrian models are narrow in scope and limited to go/no-go crossing decisions in single-lane settings. While data-driven approaches can predict pedestrian behavior in complex situations, they lack sufficient observations in rare, safety-critical scenarios. Here, we propose an approach to training pedestrian models in simulators so that learned policies generate demonstrably human-like behavior in realistic, complex traffic scenarios, including multiple lanes, heavy traffic, and dangerous driving styles. Our technical contribution is a novel definition of pedestrian-vehicle interaction as a partially observable Markov decision process (POMDP) with theory-grounded perceptual, cognitive, and motor constraints. It accounts for the highly adaptive nature of human behavior in traffic and simulates how people adjust their responses according to perceived danger, time pressure, and the complexity of the situation. When trained via deep reinforcement learning (RL) with domain randomization in a simulator, the model reproduces the broadest range of empirical findings shown so far on human crossing behavior, including gap acceptance, yielding acceptance, hesitation, and evasive speed adjustment. We show that learned policies transfer to unseen traffic environments, and can be further adapted to local traffic norms with finetuning. Together, these results establish a blueprint for simulator-ready pedestrian models that can support the development and evaluation of automated driving systems.

Problem

• pedestrian model
• automated driving simulation
• human-like behavior
• safety-critical scenarios

Innovation

• Partially Observable Markov Decision Process (POMDP)
• Deep Reinforcement Learning
• Domain Randomization
• Pedestrian-Vehicle Interaction
• Automated Driving Simulation