At around 9 a.m. in Guangzhou, Guangdong Province, Chen Li opened a mobile app and booked a fully driverless ride-hailing car. Ten minutes later, a business vehicle pulled up at the curb and carried him steadily toward his destination.
Chen Li boarded at Guangzhou East Station 鈥?a spot notorious among locals as a "hard-to-drive" area, not only because of heavy vehicle traffic, but also because lanes have been narrowed by subway construction. Yet in the busiest and most congested part of Guangzhou, a car with no driver simply "drove" off.
The convenience Chen Li felt is backed by a set of rapidly updating figures. One smart mobility company's ride-hailing fleet in Guangzhou has already achieved per-vehicle profitability 鈥?the money each car earns per day is enough to cover its own operating costs. Another smart mobility company has obtained autonomous driving licenses in nine countries, with operations covering more than 60 cities across 13 countries.
The explorations of these two enterprises are accompanied by the accelerated implementation of national strategy. Recently, nine government departments including the Ministry of Industry and Information Technology jointly issued the "15th Five-Year Plan for the Development of the Intelligent Connected New Energy Vehicle Industry," which explicitly proposes that by 2030, vehicles with autonomous driving functions will achieve large-scale application. The mandatory national standard "Safety Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles," led by the Ministry of Industry and Information Technology, has also been publicly announced and taken effect.
Under the traction of national strategy, Guangzhou's report card is as follows: as of the end of September 2026, more than 20 entities had been approved for road testing and demonstration activity permits, more than 1,000 vehicles had been deployed cumulatively, 1,375 test roads had been opened cumulatively, with a one-way mileage of approximately 2,661.58 kilometers.
"Smart" vehicles drive into city streets and alleys
Many cities are laying out autonomous driving industries. What makes Guangzhou distinctive? What left a deep impression on this reporter was the "multi-scenario commercial operation pilot" and the "working approach for exploring mixed-traffic autonomous driving pilot projects" mentioned by a relevant official from the Automobile Industry Division of the Guangzhou Municipal Bureau of Industry and Information Technology.
The "multi-scenario" approach described by that official can be found one by one in the streets and alleys of Guangzhou. Open a certain ride-hailing app, and autonomous driving already sits quietly in the function bar. Through a mixed dispatch model, it serves as supplementary capacity and is integrated into a mature platform dispatch system. On Bio-Island in Huangpu District, one smart mobility company's autonomous minibus runs every ten minutes, looping between south and north, making commuting easier for employees of hundreds of companies on the island and also making travel easier for residents who come to the island for sports and recreation. In Huangpu, Tianhe, and Haizhu, the same company's driverless mobility service vehicles shuttle through major and minor streets, offering Guangzhou residents and visitors to the city a different kind of experience.
From carrying passengers to performing work, the application scenarios for autonomous driving in Guangzhou are continuing to extend into more segmented fields. Tunnels are highly enclosed and have low visibility, creating safety hazards for traditional manual sanitation work. To eliminate these hazards, fully driverless sanitation vehicles are used for cleaning operations 鈥?washing and sweeping, dry sweeping, rear spraying, and counter-flushing. With one set of actions, the road is left clean. Freight vehicles are also running on the roads. L4-level autonomous freight vehicles have been approved to conduct "pure driverless testing" and "cargo-carrying testing" for urban freight vehicles in Guangzhou.
For these scenarios to "run," an oversight and support system covering the entire city is indispensable. In Tianhe District, Guangzhou, there is an automotive technology building. Take the elevator to the seventh floor, and inside the Guangzhou Intelligent Connected Vehicle Demonstration Zone Operation Center, all autonomous vehicles and equipment in Guangzhou are closely monitored, and any abnormality is recorded by the platform.
"The more data accumulates, the easier it is to identify common causes of abnormal states, thereby providing safety recommendations to enterprises," Wang Yue, vice president of a company engaged in autonomous driving business, told the reporter.
The Hong Kong University of Science and Technology (Guangzhou) has established a dedicated Intelligent Transportation Thrust. At the Intelligent Autonomous Driving Technology Center, the reporter saw a driving simulator with six degrees of freedom. He Dengbo, assistant professor of the Intelligent Transportation Thrust under the Systems Hub, pointed to four other types of simulation equipment and introduced them: "They can run synchronously in the same virtual environment, simulating real-time interaction among multiple road users such as drivers and pedestrians in the same scenario. By observing interactions among different traffic participants, the laboratory can analyze interaction forms, optimize interaction design, and provide data support for the formulation of laws and regulations."
A safety net woven tightly across every kilometer
Scenarios can be diverse, but the premise must be safety. Sitting in the car, Chen Li felt very reassured: the in-car screen displayed real-time road conditions and a 3D visualization of vehicle behavior, making the status clear at a glance. The car was equipped with a safety hammer and a safety notice card, and voice prompts and detection devices ensured that passengers fastened their seat belts and avoided overweight or overloading. If needed, passengers could contact customer service immediately through the in-car screen or mobile phone.
Chen Li's feeling did not come only from the visible safety features inside the car. The vehicle was equipped with nine lidars and 14 cameras, forming an overlapping perception network. "When our fully driverless vehicles run on the road, three 'brain systems' simultaneously ensure safety. The main system handles all complex road condition recognition, decision-making, and planning, just like a driver watching in all directions and listening on all sides. At the same time, there is a 'backup brain.' Once the main system fails, the backup chip intervenes," explained Mo Luyi, vice president of a smart mobility company.
In addition to the important guarantee provided by technology, government regulation has also designed a "small-step, fast-run" oversight plan. Every vehicle must go through a strict verification process before going on the road. Wu Zhengming, director of the Guangzhou Intelligent Connected Vehicle Demonstration Zone Operation Center, told the reporter that Guangzhou adopts a citywide "one chessboard" approach to autonomous driving development. Before going on the road, all vehicles must undergo autonomous driving function testing in closed venues, "similar to an entrance examination." After meeting road standards, Guangzhou divides public roads into two major categories: general test roads and high-speed and expressway test roads. Each category is further divided into levels one, two, and three, for a total of six levels. Vehicles must follow a progressive mechanism similar to a "school advancement examination." Only after accumulating the corresponding mileage without safety accidents and meeting relevant requirements for the average mileage before the autonomous driving system is taken over by a safety officer can they apply for an upgrade. If a vehicle fails assessment at a certain level or has safety problems, it cannot enter the next stage.
After a series of "examinations," Guangzhou has accumulated more than 25 million kilometers of test data and more than 40 million kilometers of operational data. Data generated on real roads is the foundation, but real-road testing alone is far from enough. "In addition to running on the road, we also have a simulation system, allowing vehicles to accumulate more experience," said Li Yifan, vice president of a smart mobility company in Guangzhou. "Compared with passively waiting for problems to appear on real roads, the simulation system can actively generate various complex scenarios, allowing low-frequency long-tail problems such as extreme weather, construction detours, and sudden cut-ins to receive full training and verification."
Supportive policies build a solid foundation for the industry
"What makes Guangzhou's autonomous driving policy advanced is not that it 'flies a kite,' but that it clearly draws the track for enterprises and escorts them all the way. Its pragmatism is reflected in three aspects: it started early, it is steady, and it is deep," Li Yifan said.
First, Guangzhou was very early in giving enterprises real roads and real scenarios for technical verification under controllable risks and overall coordination. Second, the policy is not a single-point breakthrough, but institutionalized and systematized advancement. From test licenses and demonstration operations to commercial operations, there are clear rules. Third, and what enterprises value most, the policy is truly designed around cultivating the industry. Starting from real needs such as mobility, logistics, and sanitation, it helps autonomous driving find diverse scenarios for large-scale application, enabling enterprises to truly achieve "technological productization and product commercialization."
The "early start" Li Yifan mentioned is especially evident and specific in Huangpu District. In 2017, when his company settled in Huangpu District, autonomous driving technology was just beginning to emerge. In March 2018, "driverless driving" first appeared in the Huangpu District government work report. In July 2019, Huangpu District deployed China's first 5G autonomous driving application demonstration island 鈥?Bio-Island. In the same year, it opened the city's first batch of open test roads and took the lead nationwide in launching autonomous driving bus and autonomous driving passenger vehicle services open to the public. In 2020, after being approved as one of the city's first pilot areas for intelligent connected vehicle road testing, it focused on advancing the opening of test roads. In 2024, it achieved district-wide road opening, cumulatively opening 371 test roads across three levels in eleven batches.
Behind the running of Huangpu District's "smart vehicles" is the simultaneous construction of "smart roads." At present, Huangpu District's roadside perception system has covered 332 intersections and deployed 3,497 intelligent connected perception devices. In obscured "ghost probe" scenarios, visual blind spots at intersections, pedestrians crossing suddenly, and risks of vehicles from different directions cutting in, driving in reverse, or changing lanes, vehicles no longer rely entirely on their own recognition. Roadside equipment can also warn vehicles, making up for the perception blind spots of single-vehicle sensors.
Beyond the forward-looking layout of policies, Guangzhou has a deeper foundation for developing autonomous driving 鈥?the industrial base of the automobile industry. Guangzhou has long built a complete industrial chain from core components to complete vehicle manufacturing, gathering a group of complete vehicle manufacturers, autonomous driving technology companies, and upstream and downstream enterprises in the industrial chain. It promotes coordinated development of complete vehicle manufacturing with key technologies such as autonomous driving algorithms, intelligent sensors, automotive-grade chips, and in-vehicle operating systems. This industrial chain advantage is most directly reflected in cooperation between complete vehicle enterprises and autonomous driving companies.
"Previously, most components used by L4-level driverless companies were cutting-edge, high-performance products that were not yet mature and stable, and were costly. However, with the rise of the L2 assisted driving vehicle market, many related component industries have continued to develop and mature, prices have gradually come down, and technical solutions have become more mature. Autonomous driving components are transitioning from 'cutting-edge products' to 'automotive-grade products.' Automotive-grade standards mean component performance is more stable, safer, and more cost-controllable. The cost of autonomous driving kits has dropped by more than 70% compared with the previous generation," Mo Luyi said.
Comments