Lane-deviation penalty formulation and analysis for autonomous vehicle avoidance maneuvers

Anistratov, Pavel; Olofsson, Björn; Nielsen, Lars · 2021 · Crossref

DOI: 10.1177/09544070211007979

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of designing safe and efficient avoidance maneuvers for autonomous vehicles, specifically focusing on the "moose test" or double lane-change scenario. The research is motivated by the need for autonomous systems to handle critical situations at the limit of road–tire friction, a capability required for higher levels of autonomy. The authors identify a key safety observation: while it is dangerous for a vehicle to occupy the opposing lane during an avoidance maneuver, it is safe once the vehicle returns to its original lane. To capture this dynamic, the study introduces a Lane-Deviation Penalty (LDP) objective function that penalizes the vehicle for being outside its own lane, rather than using traditional criteria like minimum time or simple lateral deviation penalties. The methodology involves formulating the motion-planning problem as an optimal control problem. The authors utilize a detailed double-track vehicle model that accounts for pitch and roll dynamics, load transfer, and wheel dynamics. Tire forces are modeled using Pacejka’s Magic Formula to represent behavior on dry asphalt. The optimization minimizes an integral cost function composed of the LDP and a recovery-behavior extension, which promotes a smooth return to the original lane after the obstacle. The problem is subject to physical constraints, including limits on steering angle, wheel torques, and road boundaries defined by the ISO 3888-2 standard. The vehicle model is implemented in Modelica, and the optimal trajectories are computed for various scenarios by varying parameters such as initial vehicle speed, obstacle size, and obstacle placement. The results demonstrate that the LDP formulation yields efficient and stable maneuvers across a range of initial velocities, from low speeds where emergency braking is possible to high speeds near the handling limit. A key finding is that the LDP approach significantly reduces the time the vehicle spends in the opposing lane compared to maneuvers optimized for minimum time or other lateral-penalty functions. Despite the LDP being a position-based penalty, it outperforms the minimum-time criterion in terms of safety exposure in the opposing lane. The study also illustrates the vehicle's behavior under different obstacle configurations, showing that the method provides robust solutions for emergency avoidance. The significance of this work lies in its contribution to the development of active safety systems for autonomous vehicles. By providing a penalty formulation that explicitly accounts for the danger of occupying the opposing lane, the LDP method offers a more safety-oriented approach to motion planning than traditional time-minimization strategies. The use of a comprehensive double-track model with load transfer adds realism to the analysis, ensuring that the proposed control strategies are viable for real-world vehicle dynamics. This research supports the advancement of autonomous driving technologies by offering a validated method for handling critical avoidance maneuvers at the limits of vehicle performance.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• lane positioning