A spatiotemporal trajectory data-driven meta-modeling method for autonomous driving safety scenarios

Abstract

The invention discloses a space-time trajectory data-driven meta-modeling method for automatic driving safety scenarios, which is characterized by constructing a meta-modeling method for space-time trajectory data based on a MOF meta-modeling system. The construction, preprocessing of spatiotemporal trajectory data, and data quality evaluation can obtain high-quality spatiotemporal trajectory data of autonomous vehicles, and then a MOF-based spatiotemporal trajectory data meta-modeling process and a spatiotemporal trajectory data meta-model for auto-driving are constructed. And automatically instantiate security scenarios based on ADSML, a scenario modeling language. The present invention proposes a meta-modeling method of automobile automatic driving safety scene based on spatiotemporal trajectory data driving, which can effectively support model-driven and data-driven collaborative development methods, and provides a new approach for scene modeling in the field of automatic driving.

Description

technical field [0001] The invention relates to a meta-modeling method for safety scenarios, in particular to a meta-modeling method for automatic driving safety scenarios driven by spatiotemporal trajectory data. Background technique [0002] In recent years, the rapid development of big data, artificial intelligence, autonomous driving, smart city, intelligent transportation and other fields has promoted the wide application of the new generation of information technology. How to efficiently utilize the large amount of data generated is an urgent problem to be solved. Effective spatiotemporal fusion, high-confidence modeling, big data processing, and spatiotemporal data applications for specific fields between the spatiotemporal data generated by earth observation and public media data (such as city cameras, social media, personal activities, etc.) Research hotspots in the field of spatial information technology and big data applications. Especially the global navigation ...

AI Technical Summary

Problems solved by technology

[0004] However, at present, the scene modeling methods and technologies for real spatio-temporal trajectory data still have deficiencies: the research on autonomous driving scene modeling mainly focuses on the design of the modeling scene, emphasizing on the simulation results, without a set of systematic models The modeling method of driver and data-driven collaborative development is inconvenient for the management of real data and the generation of high-quality data, and there are problems of lack of a readable scene description language and difficulty in formal verification (Formal Verification)

Although GeoScenario and Scenic are more concise and readable than other scene modeling languages, their description methods are not friendly enough for domain experts, and formal methods cannot be used to verify and analyze scenarios

Images