Risk Assessment on Off-road Environment from the Driver's Perspective

Reasonable risk assessment is pivotal in aiding path planning algorithms to ascertain safe, collision-free paths while minimizing the path length. In this paper, we propose a unified risk field as a method of risk assessment from the driver's perspective. It synthesizes and quantifies the relationship between environmental factors, ego vehicle characteristics, and the occupants' attributes, which include passengers and cargo, enabling a comprehensive evaluation of risks. A notable feature of our method is the integration of Non-Uniform Safety Margin Expression (NSME) into the unified risk field, enabling adaptive and anisotropic safety margins. This enhancement significantly improves vehicle driving efficiency. Additionally, by incorporating driving style constraints, our method provides a more personalized risk assessment. Moreover, our approach considers terrain-related risks to minimize path slope and enhance safety. To validate the effectiveness of our method, we integrate it with the Probabilistic Roadmap (PRM) path planning algorithm and conduct thorough simulations.

I. Introduction

The crucial role of path planning algorithms in autonomous driving necessitates a balanced consideration of safety and efficiency, with risk assessment being pivotal in achieving this balance. Risk assessment can be broadly categorized into two perspectives: those focused on traffic management and those centered on vehicle navigation. Traffic management-oriented approaches typically employ statistical indicators, Road Safety Development Index (RSDI) [1], time series analyses, and regression techniques. However, these methods often rely heavily on historical accident data, limiting their ability to address real-time road conditions effectively.