Shanghai's first training data collection vehicles with autonomous driving large model hit road

Shanghai (Gasgoo)- On December 31, 2024, a fleet of 30 IM L6 data collection vehicles was officially launched at a ceremony in Pudong New Area, marking a significant step in Shanghai's efforts to advance its high-level autonomous driving ecosystem.

As part of its vision to establish Shanghai as a hub for artificial intelligence, the city is developing a high-level autonomous driving pilot zone. This initiative follows a technology roadmap based on "vehicle intelligence as the foundation and vehicle-road-cloud collaboration as the key support."

A crucial aspect of this initiative is to carry out the collection and processing of real-world vehicle training data. Leveraging data collection vehicles from the ride-hailing platform, automakers, and autonomous driving technology companies, Shanghai is gathering these data and applying a closed-loop data service toolchain involving collection, cleansing, labeling, testing, and application to build a massive dataset and scenario library supporting large-scale autonomous driving large model training.

In addition, the city aims to construct a training platform for autonomous vehicles. By integrating citywide vehicle-collected training data and real-time roadside data like traffic signal data and traffic information, this platform is set to create virtual dataset segments with diverse scenarios through world models, supporting full-scenario model training and closed-loop simulation evaluations.

Furthermore, Shanghai is set to establish a vehicle-to-everything (V2X) communication network. At the initial stage, a 100-kilometer 5G-A V2X network will be deployed in the Pudong New Area, enabling closed-loop verification of typical application scenarios.

The first batch of 30 data collection vehicles, which are operated as ride-hailing vehicles by SAIC Mobility, the mobility service brand of China's largest carmaker SAIC Motor, will gather high-quality driving scenario data from skilled drivers, significantly increasing the volume of collected data. Additionally, data from 70 existing corporate-owned data collection vehicles has been integrated into the shared data pool, allowing Shanghai to achieve its goal of deploying 100 data collection vehicles on roads this year.