General virtual sensor data collection method and system for virtual algorithm verification

Abstract

A general virtual sensor model modeling method oriented to virtual algorithm verification. By establishing model indicators and adopting a new modeling technology based on MCMC method, a real distribution model is established, and a virtual sensor model that can be included in a virtual simulation platform is established. The operation of the data collection stage is simplified, and the classification and processing of data can be realized automatically, improving the efficiency of data collection. The virtual sensor model established by the present invention can reflect the detection probability, classification ability, and motion state detection accuracy of the sensor under the specific target state conditions in the real environment, endow the virtual sensor with the failure ability in line with the real situation in the virtual test, and facilitate the verification of planning and control algorithms Robustness and effectiveness in extreme scenarios such as sensor failure.

Description

technical field [0001] The present invention relates to a vehicle-mounted sensor virtual simulation technology in the field of automatic driving, in particular to a general virtual sensor modeling method and system for verification of intelligent networked vehicle planning control algorithms, applicable to different types of exogenous sensor devices. Background technique [0002] The testing of smart car algorithms needs to cover enough scenarios, especially those with high hazards. Based on real vehicle experiments, it is impossible to completely cover the most harmful test scenarios, so virtual testing has become the mainstream form to meet this challenge. The virtual sensor in the existing virtual test software cannot reflect the characteristics of the real sensor, specifically, the error of the sensor in the detection of the surrounding perception target caused by the complexity of the natural traffic environment cannot be reflected. For example, the missed detection an...

AI Technical Summary

Problems solved by technology

Currently based on sensor physical model modeling (such as the sensor model in IPG CarMaker / PreScan), although it is convenient to establish virtual sensor models of different models and parameters, it cannot model the interference of real complex environments on measurement

The simulation of automatic driving algorithm / ADAS algorithm based on the current sensor model will not fully reflect the possible missed detection, false detection, detection error, etc. of the sensor, and cannot fully verify the robustness of the automatic driving algorithm / ADAS algorithm

Images