Risk Estimation for Automated Driving

TL;DR

This work tackles risk estimation for automated driving by combining collision probability with collision severity through a multi-circular footprint that over-approximates vehicle shapes. The authors extend a prior collision-probability framework to risk by associating per-circle severities and aggregating them across collision constellations, enabling real-time computation. They adopt Gaussian position uncertainties and a wrapped heading model, using a kinetic-energy-based severity that weights different circle-to-circle collisions, and derive a tractable two-dimensional integration that can be evaluated efficiently. Numerical case studies across head-on, rear-end, and side collisions demonstrate accurate risk estimation with practical computation times, and the approach is released as open-source for real-time motion-planning applications.

Abstract

Safety is a central requirement for automated vehicles. As such, the assessment of risk in automated driving is key in supporting both motion planning technologies and safety evaluation. In automated driving, risk is characterized by two aspects. The first aspect is the uncertainty on the state estimates of other road participants by an automated vehicle. The second aspect is the severity of a collision event with said traffic participants. Here, the uncertainty aspect typically causes the risk to be non-zero for near-collision events. This makes risk particularly useful for automated vehicle motion planning. Namely, constraining or minimizing risk naturally navigates the automated vehicle around traffic participants while keeping a safety distance based on the level of uncertainty and the potential severity of the impending collision. Existing approaches to calculate the risk either resort to empirical modeling or severe approximations, and, hence, lack generalizability and accuracy. In this paper, we combine recent advances in collision probability estimation with the concept of collision severity to develop a general method for accurate risk estimation. The proposed method allows us to assign individual severity functions for different collision constellations, such as, e.g., frontal or side collisions. Furthermore, we show that the proposed approach is computationally efficient, which is beneficial, e.g., in real-time motion planning applications. The programming code for an exemplary implementation of Gaussian uncertainties is also provided.

Figures (5)

• Figure 1: Problem Statement: The ego (blue vehicle) is colliding with its front section into the object's (white vehicle) front and side section colored red and orange, respectively.
• Figure 2: Example: Intersection angle intervals. In all subfigures, the object (white vehicle) is located at the same relative position $(\phi, \rho)$ to the ego (blue vehicle).
• Figure 3: Illustration of Example \ref{['ex:intersection_intervals']}: Construction of disjoint intersection angle intervals.
• Figure 4: Cases (I): the object (dark blue vehicle) is colliding with the ego (light blue vehicle) head-on. Case (II): The ego collides into the object's (red vehicle) rear.
• Figure 5: Side collision Cases (III) - (V), the ego is displayed in light blue and the object's color is different in each case.

Theorems & Definitions (3)

• Example 1
• Example 2
• Remark 1