Tackling Occlusions & Limited Sensor Range with Set-based Safety Verification

Orzechowski, Piotr F.; Meyer, Annika; Lauer, Martin · 2018 · OpenAlex-citations

DOI: 10.1109/itsc.2018.8569332

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 ensuring provable safety for automated vehicles in environments characterized by occlusions and limited sensor ranges. While recent approaches utilize probabilistic methods or machine learning to manage risk under uncertainty, they often fail to guarantee safety. The authors argue for a methodology that prioritizes safety verification over comfort, proposing a set-based approach to verify trajectories against arbitrary safe-state formulations even when obstacle states are unknown. The method formalizes potential risk by identifying critical sensing field edges—borders of the ego vehicle’s field of view that could hide obstacles. For each critical edge, the authors define virtual obstacles with initial states represented as intervals covering all physically possible positions, orientations, and velocities. They extend existing reachable set over-approximation techniques (specifically M1 for acceleration-based occupancy and M2 for lane-following occupancy) to handle these interval-based initial states. This allows the system to compute conservative occupancy predictions for potentially hidden obstacles. A trajectory planner then verifies if a candidate trajectory, including a fail-safe maneuver option, remains collision-free with respect to these over-approximated reachable sets. If verification fails, the trajectory is iteratively adapted until safety is proven. The approach was evaluated in three simulative scenarios. In a merging scenario, the ego vehicle successfully reduced its velocity upon detecting that its intended trajectory intersected the occupancy prediction of a potentially occluded obstacle. When an obstacle actually appeared from behind the occlusion, the vehicle executed a safe stop. In other scenarios involving intersections and limited sensor ranges, the method demonstrated the ability to adapt trajectories to guarantee safety despite incomplete environment knowledge. The results show that the planner can maintain safe behavior by relying on the verified fail-safe options rather than accurate intention estimation of other agents. The significance of this work lies in its contribution to formal safety verification for automated driving systems. By providing a method to over-approximate the reachable sets of occluded obstacles, it enables the verification of trajectories under high uncertainty. This approach supports the development of automated vehicles that can legally and safely operate in complex traffic scenarios, offering a rigorous alternative to risk-minimization strategies that lack formal safety guarantees. The method is applicable to various levels of automation and integrates easily with existing uncertainty models.

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.

• situational awareness