An improved 2D time-to-collision for articulated vehicles: predicting sideswipe and rear-end collisions

2025-07-05

AI Summary

Conventional two-dimensional time-to-collision (TTC₂D) models assume constant vehicle heading, speed, and acceleration—assumptions that fail to capture complex collision dynamics of articulated vehicles (e.g., tractor-trailers), particularly simultaneous sideswipe and rear-end collisions. Method: This paper proposes three enhanced TTC₂D formulations: (i) incorporating dynamic heading angle evolution into TTC₂D modeling; (ii) extending the framework to articulated vehicle configurations; and (iii) relaxing the constant-acceleration assumption to enable joint quantification of lateral and longitudinal collision risks. Contribution/Results: Evaluated in representative cut-in scenarios using the CARLA simulation platform, the proposed methods simultaneously detect both rear-end and sideswipe collision risks—overcoming the fundamental limitation of traditional TTC₂D, which supports only single-mode collision prediction. Experimental results demonstrate significantly improved accuracy and applicability of collision risk assessment in complex, realistic traffic scenarios.

Technology Category

• Planning, Routing, and Scheduling: Model-Based Reasoning
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Time-to-collision (TTC) is a widely used measure for estimating the time until a rear-end collision between two vehicles, assuming both maintain constant speeds and headings in the prediction horizon. To also capture sideswipe collisions, a two-dimensional extension, TTC$_{ ext{2D}}$, was introduced. However, this formulation assumes both vehicles have the same heading and that their headings remain unchanged during the manoeuvre, in addition to the standard assumptions on the prediction horizon. Moreover, its use for articulated vehicles like a tractor-semitrailer remains unclear. This paper addresses these limitations by developing three enhanced versions of TTC$_{ ext{2D}}$. The first incorporates vehicle heading information, which is missing in the original formulation. The standard assumption of constant speed and heading in the prediction horizon holds. The second adapts this to articulated vehicles while retaining the assumptions of the first version. The third version maintains the constant heading assumption but relaxes the constant speed assumption by allowing constant acceleration. The versions are tested in a cut-in scenario using the CARLA simulation environment. They detect rear-end collisions, similar to TTC, and moreover, they also identify sideswipe risks, something TTC could not predict.

Problem

• Extends 2D time-to-collision to predict sideswipe and rear-end collisions
• Enhances TTC2D for articulated vehicles like tractor-semitrailers
• Relaxes constant speed assumption by incorporating constant acceleration

Innovation

• Incorporates vehicle heading in TTC2D
• Adapts TTC2D for articulated vehicles
• Allows constant acceleration in prediction