Test scene automatic generation method, device, equipment and storage medium

Abstract

Embodiments of the present invention provide a method, device, device, and storage medium for automatically generating test scenarios, and relate to the technical field of automatic driving simulation testing. The method includes: collecting driving data in real scenes; dividing two-dimensional parameter ranges in each group of two-dimensional space data; dividing each divided group of two-dimensional parameter ranges again so as to obtain dangerous scene data and boundary scene data respectively and safety scene data; the dangerous scene data divided from each group of two-dimensional parameter ranges are synthesized into a dangerous scene space, and the boundary scene data divided from each group of two-dimensional parameter ranges are synthesized into a boundary scene space; The safety scene data divided in the dimension parameter range is synthesized into a safety scene space; the test scene data is obtained by sampling. The embodiment of the present invention can scientifically and accurately automatically generate three types of test scene data of danger, boundary and safety according to the driving data in the real scene.

Description

technical field [0001] The invention relates to the field of automatic driving simulation testing, and in particular, to a method, device, device and storage medium for automatic generation of test scenarios. Background technique [0002] As a field with the most research potential today, autonomous driving is playing an increasingly important role in traffic safety and other aspects. [0003] Autonomous driving tests are generally divided into simulation tests, tests in a closed field environment, and real-world tests on real roads. Although the real-world test can more representatively verify the safety of the vehicle, due to the extremely low probability of traffic accidents, even if the vehicle travels on tens of thousands of kilometers of roads, the probability of encountering extreme traffic accidents is still very small. The operations and results of the t are therefore unpredictable, and these extreme operations (or the detailed parameters that cause them) are very ...

AI Technical Summary

Problems solved by technology

Although real-world testing can more representatively verify the safety of vehicles, due to the extremely low probability of traffic accidents, even if the vehicle is driving on tens of thousands of kilometers of roads, the probability of encountering extreme traffic accidents is still very small. Therefore, the operation and results of it become unpredictable, and these extreme operations (or the detailed parameters that cause extreme operations) are very important in the training of vehicles

Furthermore, the test scenarios of the real-world test are limited, and the high test cost makes it difficult to cover various operation steps, and the special conditions and scenarios of many traffic accidents cannot be accurately restored

Images