Quantifying the Individual Differences of Driver' Risk Perception with Just Four Interpretable Parameters
URL: http://arxiv.org/abs/2211.10907v1
archive: archived pipeline: cataloged verified
Abstract
There will be a long time when automated vehicles are mixed with human-driven vehicles. Understanding how drivers assess driving risks and modelling their individual differences are significant for automated vehicles to develop human-like and customized behaviors, so as to gain people's trust and acceptance. However, the reality is that existing driving risk models are developed at a statistical level, and no one scenario-universal driving risk measure can correctly describe risk perception differences among drivers. We proposed a concise yet effective model, called Potential Damage Risk (PODAR) model, which provides a universal and physically meaningful structure for driving risk estimation and is suitable for general non-collision and collision scenes. In this paper, based on an open-accessed dataset collected from an obstacle avoidance experiment, four physical-interpretable parameters in PODAR, including prediction horizon, damage scale, temporal attenuation, and spatial attention, are calibrated and consequently individual risk perception models are established for each driver. The results prove the capacity and potential of PODAR to model individual differences in perceived driving risk, laying the foundation for autonomous driving to develop human-like behaviors.
Summary
Modeling study (arXiv 2022) calibrating the Potential Damage Risk (PODAR) driving-risk model individually for eight drivers from an open obstacle-avoidance simulator dataset. PODAR represents driving risk as a projection of future potential collision damage and is parameterized by four physically interpretable quantities: prediction horizon, damage scale, temporal attenuation, and spatial attention. Individual parameter sets are fit per driver and compared to characterize differences in risk perception relevant to human-like AV behavior.
Key finding
Just four interpretable PODAR parameters (prediction horizon, damage scale, temporal attenuation, spatial attention) suffice to capture per-driver differences in risk perception, enabling customizable, human-like risk assessment for autonomous-vehicle decision and planning without abandoning a physically meaningful structure.
Methodology
Calibration of the PODAR model on an open obstacle-avoidance simulator dataset; per-driver parameter fitting via optimization on observed risk responses; comparison of individual parameter sets across eight drivers.
Sample size: N=8 drivers (re-analysis of an open simulator dataset)
Quality score: 5 / 5