上海公布重磅文件:量产L4级汽车,拓展自动驾驶开放区域
Di Yi Cai Jing·2026-01-15 09:24

Core Viewpoint - Shanghai aims to accelerate the construction of a leading area for high-level autonomous driving by fostering industry-leading autonomous driving models and implementing the "Mosu Zhixing" action plan by 2027 [2][3] Group 1: Action Plan Overview - The action plan outlines the goal of achieving large-scale implementation of high-level autonomous driving applications by 2027, forming a competitive smart connected vehicle industry cluster [2] - The plan includes the exploration of innovative business models for L4-level autonomous driving technology in scenarios such as smart buses, taxis, and heavy trucks, with an open area of 2000 square kilometers and over 5000 kilometers of roads designated for autonomous driving [2][3] - The production ratio of new vehicles with combination driving assistance (L2) and conditional autonomous driving (L3) features is expected to exceed 90%, with L4 autonomous vehicles achieving mass production [2] Group 2: Infrastructure and Testing - Shanghai has opened 3173 testing roads with a total length of 5238.82 kilometers, covering about one-third of the city's area, with full-area openness in the Pudong New Area [3] - The city plans to gradually expand the autonomous driving open areas, focusing on key locations such as the Hongqiao hub, Pudong Airport, and Disneyland [2][3] Group 3: Data and Technology Development - The autonomous driving industry is currently in a "data-driven" phase, facing challenges such as high costs and difficulties in obtaining training data, insufficient scene distribution, and inadequate model testing [4] - Shanghai is promoting data sharing applications to support the training and iteration of benchmark intelligent driving models, including the deployment of data collection vehicles and the establishment of a secure data sharing mechanism [4] - A digital twin training ground for autonomous driving will be built, utilizing both real data collection and virtual generation to accumulate high-quality training datasets [4]