GB 44721—2026

The national standard lists complex meteorological conditions including rain, fog, strong light, night environment and backlight as mandatory inspection items for perception systems. However, outdoor natural weather is highly random, working conditions cannot be stably reproduced, and real road test data cannot be adopted for compliance certification.

Digauto independently developed the VTEHIL (Vehicle-Traffic-Environment Hardware-in-the-Loop) test system, which can accurately quantify and adjust composite environments such as fog density, rainfall intensity, illumination intensity and backlight, and stably reproduce all harsh scenarios required by the national standard indoors.

Full tests on key national standard items such as AEB can be carried out on the VTEHIL bench. The overall test efficiency is 2 to 3 times higher than that of outdoor real road tests, with more prominent advantages in large-scale scenario regression testing, which fully meets automakers’ demands for high-frequency iterative verification.

VTEHIL has been applied in China Automotive Engineering Research Institute laboratories with highly consistent test data, perfectly making up for the deficiencies of environmental laboratories owned by vehicle manufacturers and inspection institutions. It has served well-known universities including Tsinghua University, authoritative inspection organizations such as China Automotive Engineering Research Institute, and mainstream OEMs such as FAW Group.

DigGenX is a self-developed data generation model of Digauto. It takes real physical data collected from VTEHIL laboratories as anchor points, and generates large-scale high-fidelity multi-scenario training data through feature extraction, condition construction, world model generation and fusion post-processing, fundamentally narrowing the confidence gap between simulation and real vehicle tests. The generated data can be directly used for training and verification of end-to-end models and VLA (Vision-Language-Action) models, as well as parameter configuration and intelligent generalization of vehicle hardware-in-the-loop test scenarios.

This capability brings an in-depth change: simulation and real vehicle testing are no longer two isolated systems, but an integrated whole for mutual verification. The same test scenario can be deduced on a large scale in the simulation environment built by DigGenX and reproduced on real vehicles via the VTEHIL physical hardware-in-the-loop bench, enabling direct comparison and verification between simulation results and real vehicle performance, thus providing quantitative answers to the reliability of simulation-verified functions on real vehicles. Simulation results can be verified by real vehicle tests, and real vehicle data can be used to calibrate simulation models in reverse. Bidirectional calibration continuously reduces virtual-real deviation, making simulation confidence traceable and evidence-based, and lowering the marginal cost of each real vehicle verification.

VTEHIL and DigGenX form a closed-loop dual engine of virtual-physical integration: physical testing generates real data, real data feeds model training, and upgraded models return to physical environments for verification. This positive cycle can continuously improve the operational stability of automated driving systems under harsh environments and failure conditions, complying with the performance baseline requirements of the national standard. Meanwhile, it accumulates standardized and traceable full test data to support enterprise safety archiving, OTA iterative retesting and multi-dimensional pre-launch confirmation tests. It continuously supplies high-value training data at a cost far lower than real-road data collection, helping automakers achieve more comprehensive testing, faster iteration and lower expenditure.

We have built two standardized PGs in Yancheng and Wuzhen. Our engineering team has rich experience in site and road testing for L2-L5 automated driving, fully complying with GB/T 41798-2022 Intelligent and Connected Vehicles — Site Test Methods and Requirements for Automated Driving Functions. We provide professional and comprehensive services including ADAS R&D verification, standard compliance testing, and customized working-condition test equipment engineering.