![[2510.12206] Controllable Collision Scenario Generation via Collision Pattern Prediction](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.12206] Controllable Collision Scenario Generation via Collision Pattern Prediction
Summary
This paper introduces a novel method for generating controllable collision scenarios for autonomous vehicles, enhancing safety evaluations through simulated environments.
Why It Matters
As autonomous vehicles become more prevalent, ensuring their safety is critical. This research addresses the challenge of generating diverse and realistic collision scenarios, which are essential for testing and improving vehicle safety systems without the risks associated with real-world testing.
Key Takeaways
- Introduces controllable collision scenario generation for AVs.
- Presents COLLIDE, a dataset for diverse collision types and time-to-accident intervals.
- Demonstrates improved performance over existing methods in generating collision scenarios.
- Highlights the importance of scenario generation for enhancing AV planner robustness.
- Provides access to additional generated scenarios for further research.
Computer Science > Robotics arXiv:2510.12206 (cs) [Submitted on 14 Oct 2025 (v1), last revised 22 Feb 2026 (this version, v3)] Title:Controllable Collision Scenario Generation via Collision Pattern Prediction Authors:Pin-Lun Chen, Chi-Hsi Kung, Che-Han Chang, Wei-Chen Chiu, Yi-Ting Chen View a PDF of the paper titled Controllable Collision Scenario Generation via Collision Pattern Prediction, by Pin-Lun Chen and 4 other authors View PDF HTML (experimental) Abstract:Evaluating the safety of autonomous vehicles (AVs) requires diverse, safety-critical scenarios, with collisions being especially important yet rare and unsafe to collect in the real world. Therefore, the community has been focusing on generating safety-critical scenarios in simulation. However, controlling attributes such as collision type and time-to-accident (TTA) remains challenging. We introduce a new task called controllable collision scenario generation, where the goal is to produce trajectories that realize a user-specified collision type and TTA, to investigate the feasibility of automatically generating desired collision scenarios. To support this task, we present COLLIDE, a large-scale collision scenario dataset constructed by transforming real-world driving logs into diverse collisions, balanced across five representative collision types and different TTA intervals. We propose a framework that predicts Collision Pattern, a compact and interpretable representation that captures the spatial configuratio...