A method, system, device, and medium for co-perception risk test timing scenario generation

By analyzing natural driving datasets and generating time-series risk scenarios using parameterized models, this approach addresses the lack of temporal dynamics and collaborative perception features in existing autonomous driving tests. It enables the efficient generation of logically sound and dynamically realistic test scenarios, supporting the performance verification of collaborative perception algorithms.

CN122450840APending Publication Date: 2026-07-24TSINGHUA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TSINGHUA UNIVERSITY
Filing Date
2026-05-15
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing autonomous driving testing technologies lack temporal dynamism, physical rationality, and collaborative perception features when generating and evaluating test scenarios. In particular, the spatiotemporal information fusion problem under occlusion conditions leads to high testing costs and low coverage, making it impossible to effectively verify the performance of algorithms in temporally continuous scenarios.

Method used

By analyzing the natural driving dataset, a parameterized model is established to generate single-frame risk representation scenarios. Combined with iterative updates, temporal risk representation scenarios are generated. Temporal validity detection and occlusion constraint mechanisms are introduced to ensure the dynamic evolution and physical rationality of the scenarios.

Benefits of technology

It enables efficient generation of dynamic risk scenarios, supports performance verification of collaborative perception algorithms under complex dynamic conditions, improves testing efficiency and the logical rationality and dynamic realism of scenario generation, and is compatible with multiple simulation platforms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122450840A_ABST
    Figure CN122450840A_ABST
Patent Text Reader

Abstract

The present application relates to a kind of collaborative perception risk test timing scene generation method, system, equipment and medium, comprising: the natural driving data set obtained is analyzed, and the distribution of key parameters is modeled, obtain the parameterized model for sampling;Based on risk function scene definition, the initial motion state of key participant is sampled design, generates single frame risk concrete scene;Based on the single frame risk concrete scene generated, other participants are updated iteratively in combination with the motion state of parameterized model and key participant, obtain time-sequenced risk concrete scene;Effectiveness detection is carried out to time-sequenced risk concrete scene, ensure that each frame in time-sequenced risk concrete scene meets preset requirement;Based on the time-sequenced risk concrete scene after effectiveness detection, data rendering is carried out using simulation platform, form collaborative perception time-sequenced risk test data set.The present application can be widely applied to collaborative perception test scene generation field.
Need to check novelty before this filing date? Find Prior Art