From the Racetrack to the Road: Real-Time Trajectory Replanning for Autonomous Driving

Subosits, John; Gerdes, J. Christian · 2019 · OpenAlex-citations

DOI: 10.1109/tiv.2019.2904390

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 real-time trajectory replanning for autonomous vehicles operating near their friction limits during emergency situations, such as sudden obstacle avoidance. While autonomous vehicles typically operate safely within standard limits, emergency scenarios require maximizing vehicle dynamics to avoid collisions. Existing nonlinear optimization methods are computationally expensive and unsuitable for the fast reaction times required in these high-dynamic situations. The authors propose a scheme derived from offline racing line optimization that simultaneously modifies the vehicle’s path and speed profile in response to environmental changes, such as obstacles or reduced road friction. The method formulates the trajectory optimization problem as a quadratically constrained quadratic program (QCQP), enabling solution times of less than 20 milliseconds even with a 10-second planning horizon. The approach utilizes a simplified point mass model but enhances accuracy by incorporating longitudinal weight transfer and road topography effects, which significantly influence acceleration limits. The optimization minimizes travel time over a fixed-distance horizon, treating time as a state variable. Constraints include vehicle dynamics, road boundaries, engine power limits, and tire friction limits modeled via friction circles. The algorithm assumes the existence of a nominal trajectory and a low-level controller capable of tracking the replanned output. It reparameterizes the trajectory into a path and speed profile, converting dynamics into a spatial description to handle three-dimensional road geometries, including pitch and roll. Experimental validation was conducted using an autonomous vehicle in two scenarios: one inspired by racing dynamics and another involving aggressive obstacle avoidance. The results demonstrate that the planner successfully enables the vehicle to avoid obstacles even when they appear suddenly and the vehicle is already operating near its friction limits. The QCQP formulation allows for real-time operation with longer planning horizons compared to existing model predictive control approaches, which often use shorter horizons or decouple longitudinal and lateral planning. The inclusion of road topography and weight transfer provides a more physically motivated model of acceleration limits than comparable kinematic models, which are often invalid when maneuvers require more than half of the available tire friction. The significance of this work lies in bridging the gap between offline racing line optimization and real-time autonomous driving safety. By achieving fast computation times with a high-fidelity model of friction limits, the method allows autonomous vehicles to safely exploit their full dynamic potential in emergency scenarios. This approach improves upon previous methods by offering shorter computation times, longer planning horizons, and more accurate modeling of vehicle constraints, thereby enhancing the safety and reliability of autonomous systems in critical situations.

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.

• automation surprise