Two-dimensional time-to-collision measures for articulated vehicles: predicting sideswipe and rear-end collisions

Abstract:Time-to-collision is a commonly employed measure for rear-end collision prediction. However, its conventional formulation, which assumes constant speed and heading, is incapable of identifying sideswipe collisions. A two-dimensional extension has been proposed to incorporate lateral interactions with passenger cars, yet it assumes identical, fixed headings and does not accommodate articulated vehicles such as tractor-semitrailers. In this paper, the existing formulation for the car is first refined to incorporate differences in vehicle heading. Subsequently, new two-dimensional time-to-collision measures are proposed for articulated vehicles: TTC$_{\mathrm{2D}}^{\mathrm{AV}}$ and modified TTC$_{\mathrm{2D}}^{\mathrm{AV}}$. These measures employ constant-speed and constant-acceleration assumptions and are analogous to their one-dimensional counterparts: time-to-collision and modified time-to-collision. The proposed measures are assessed in CARLA using randomly generated cut-in scenarios simulated with a tractor-semitrailer model, incorporating a range of trailer lengths. A short analysis is also conducted to test the measures in a roundabout and tight turn. The analyses demonstrate that the proposed measures substantially improve the detection of sideswipe collisions while maintaining a comparable level of performance to existing measures in detecting rear-end collisions. Across 30 simulated scenarios, they correctly identify 14 of 15 sideswipe collisions, compared with 7 of 15 identified by the existing formulation. Moreover, the mean prediction error for sideswipe collisions is reduced by approximately 20% compared to the existing formulation.

Robotics (cs.RO)

arXiv:2507.04184 [cs.RO]

(or arXiv:2507.04184v3 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2507.04184 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.