The car becomes an "old driver"
2024-08-27
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
I am watching the research on new energy and intelligent connected vehicles on site
Original title: Cars become "old drivers"
Hubei Daily Omnimedia reporter Zuo Chen, intern Guan Shiyu, correspondent Gu Shengwei
At 3 pm on August 21, outside the Dongfeng Motor Global Innovation Center in Wuhan Economic and Technological Development Zone, Cheng Liangzhu and his colleagues were testing the car's intelligent driving performance.
Inside the invisible body of the car, there are densely packed sensors, wiring harnesses, controllers, etc. When the vehicle is driving, driver information, road information, movement of surrounding obstacles and other data are transmitted to the computer in real time. Cheng Liangzhu and his colleagues have to do such tests at least 1,000 times a week.
Researchers are calibrating the "Flash" mobile street view measurement vehicle. (Photo by Wei Lei, a reporter from Hubei Daily)
Cheng Liangzhu is mainly responsible for the development of Dongfeng Motor's intelligent driving software. The Software Engineering Research Center where he works is one of the seven centers established after Dongfeng Motor established its R&D headquarters in November 2023. At present, Cheng Liangzhu and his colleagues are working on the L2.9 advanced intelligent driving assistance system technology.
In June this year, the Dongfeng Mengshi 917 Jiaolong Armor equipped with Dongfeng's autonomous intelligent driving assistance system went on pre-sale.
"What can the L2.9 advanced intelligent driving assistance system achieve?" the reporter asked.
"Realize point-to-point intelligent driving in the city and cross-floor memory parking." Cheng Liangzhu painted a scene like this: when you turn on the intelligent driving mode with one click, the car can start, brake, overtake, and turn flexibly and smoothly, avoid vehicles and crowds easily, and can also autonomously find and park parking spaces in multi-story parking lots.
To realize this scenario, a set of AI algorithms must be created.
Cheng Liangzhu has two computers on his desk, one for viewing documents and the other for development. He told reporters that the massive data collected during the test cannot be used directly, but must be "cleaned, distilled, and labeled". Of the more than 1,000TB of raw data, only a dozen TB are available after cleaning. These data are then imported into the AI algorithm, and the algorithm is trained and optimized by comparing the control instructions of the intelligent driving system and the human driver, finally turning the car into an "experienced driver".
"Processing data and developing algorithms is very boring." Cheng Liangzhu admitted that algorithm training takes at least three to five days and at most more than a month, and requires at least tens of millions of training times.
Competition in the new energy vehicle industry is intensifying, smart connected vehicles and high-level autonomous driving are experiencing rapid development, and a vast blue ocean is right before our eyes.
Cheng Liangzhu said that independent control of core technologies is the key to high-quality development of the automobile industry. This year, Dongfeng Motor will continue to integrate R&D resources, with R&D investment exceeding 50 billion yuan within three years and R&D investment intensity of no less than 6% in 2025.
Source: Hubei Daily
