Fail-safe motion planning of autonomous vehicles

Magdici, Silvia; Althoff, Matthias · 2016 · OpenAlex-citations

DOI: 10.1109/itsc.2016.7795594

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 critical challenge of guaranteeing safety in autonomous vehicle motion planning within dynamic environments. While existing methods often generate optimal trajectories based on the most probable behavior of surrounding traffic, they fail to account for unexpected maneuvers that could lead to inevitable collisions. The authors propose a fail-safe motion planner that simultaneously ensures comfort through optimal path generation and guarantees safety by maintaining a pre-computed emergency maneuver. This approach ensures that the host vehicle can always bring itself to a standstill without colliding with other participants, regardless of their actions. The methodology employs a three-step process using a bicycle model for vehicle dynamics. First, the system predicts the most likely trajectory of leading vehicles using a Maneuver Recognition Module (MRM), which identifies intent based on lane positioning. Second, an optimal trajectory for the host vehicle is generated over a time horizon $Th_1$ using optimal control, minimizing deviation from the lane center while avoiding the predicted path of the lead vehicle. Third, and crucially, an emergency maneuver is computed over a shorter horizon $Th_2$. This maneuver is designed to avoid an overapproximative occupancy set that encloses all possible trajectories of the leading vehicle, derived from abstracted models constrained by traffic rules and physical limits. At each time step, the system checks if a new feasible optimal trajectory exists; if not, it executes the emergency maneuver to stop safely. The approach was evaluated using real traffic data from the Federal Highway Administration’s Next Generation Simulation (NGSIM) project, collected on the U.S. 101 Highway. Three scenarios were simulated. In the first scenario, where the planner relied solely on the most probable trajectory of the lead vehicle, an unexpected lane change by the lead vehicle resulted in a collision. In the second scenario, applying the proposed fail-safe planner allowed the host vehicle to detect the unexpected maneuver and execute an emergency braking and steering action, successfully avoiding the crash. The third scenario involved two leading vehicles; the host vehicle successfully avoided a collision when one vehicle unexpectedly entered its lane and continued driving safely when the other vehicle aborted a lane change. The significance of this work lies in its ability to formally guarantee safety without sacrificing comfort during normal operation. By decoupling the optimal trajectory from the safety guarantee, the planner provides a robust fallback mechanism that accounts for the infinite possibilities of other drivers' behaviors. The use of abstracted models for occupancy prediction ensures computational efficiency suitable for real-time applications. This framework addresses a major gap in autonomous driving research by providing a mathematically verified method to handle uncertainty and unexpected maneuvers, thereby enhancing the reliability of autonomous systems in complex traffic environments.

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