Every reporter Huang Xinxu and every editor Sun Lei
Recently, a video of "Zhiji LS6 APA automatic parking scraping I-shaped metal columns" has aroused heated discussions.
According to the video, during the automatic parking of a Zhiji LS6, the front of the passenger side collided with the pillar of the parking lot, but neither the vehicle system nor the owner of the Zhiji found any abnormality. Eventually, when the front bumper was smashed, the vehicle automatically parked into the parking space.
After the incident, Wang Kang, project director of Zhiji Auto's autonomous driving center, posted on Weibo that he, as the person in charge of the Zhiji Autonomous Driving team, was willing to bear the customer's maintenance costs at his own expense, and said that after the Zhiji IMAD team discovered the accident, it immediately launched a high-quality analysis of the case, proposed technical solutions, and took improvement measures. IM Zhiji Auto's official Weibo responded that the company has negotiated with users to deal with the plan.
Image source: Sina Weibo @IM Zhiji-Wang Kang
"In this accident scene, the thin column is a 'I'-shaped metal structure, and the size is small. As a result, the accuracy of the obstacle position constructed by the perception sensors such as the vehicle's ultrasonic radar and camera is insufficient, resulting in the collision between the vehicle and the metal pillar. Zhiji Auto analyzed the accident.
Generally speaking, from a technical point of view, behind intelligent parking is a complex system, mainly through ultrasonic radar, visual perception and lidar (if any) and other hardware to help vehicles accurately obtain parking spaces, obstacles, pedestrians and other information, but also inseparable from algorithms and other software systems to achieve efficient data processing and information fusion, collision avoidance decisions.
"For all kinds of obstacles in the parking scene, the basic principle of recognition is as follows: on the one hand, vision constructs environmental information by identifying the grounding wires of pillars, walls, curbs and ground; On the other hand, the environmental information is constructed through radar echo synchronization, and finally the perception results are output through model fusion. For obstacles with small diameter, short height, sharp edges, hanging or irregular shapes, the radar echo may not be accurate or no echo, resulting in the inaccurate location and size of the detected obstacles, and the environmental obstacle information constructed by the final fusion result is biased. The staff of Zhiji Automobile told the "Daily Economic News" reporter.
According to the reporter's understanding, there are generally some weak scenarios in the automatic parking industry. For example, thin column type obstacles; low steps, low stone piers, benches, etc.; Hanging objects, hanging pipes, hanging fire boxes, ventilation pipes, iron door handles, hanging flower pots and other protruding objects, etc.; Forklifts, buckets, forklifts, bulldozers, pallet trucks, cranes, extra-long transport trucks and other special-shaped projects, envelope vehicles, etc.; Ground pits, drainage ditches and other pits and other scenes, ground raised obstacles, sundries and other scenes, as well as mechanical parking spaces, three-dimensional parking spaces.
Vulnerable scene of automatic parking (Image source: Courtesy of the interviewee)
Vulnerable scene of automatic parking (Image source: Courtesy of the interviewee)
"When parking, dodging alien obstacles is a common problem. At this stage, the Visual Fusion Parking Assist function is not automatic parking, and the driver still needs to pay attention to the safety of the surrounding environment and be ready to take over the vehicle at any time to be safe. An industry source told reporters.
In addition, the reporter also consulted the instructions on the use of the "one-button parking" and "parking assist system" functions in the Zhiji LS6 user manual on the official website of Zhiji Auto. Among them, it is mentioned that thin pillars and small objects may be difficult to detect by ultrasonic radar, and the driver should judge whether to use the system according to the overall environment (such as weather, geographical environment, etc.).
Image source: Zhiji Auto's official website
Image source: Zhiji Auto's official website
On May 17, Wang Kang revealed on Weibo that Zhiji Auto's intelligent driving team went to the scene of the accident in Wuhan a few days ago to conduct an APA parking retest. "After more than 60 field tests, the Zhiji LS6 has completed obstacle avoidance and parking storage. Based on yesterday's all-day test, the optimization of the parking function has basically met the expected requirements, and it will be equipped with IMOS 2.8.0 system in the future, and will be pushed to users together with other optimized content. Wang Kang said.
Image source: Sina Weibo @IM Zhiji-Wang Kang
From the beginning of its birth, Zhiji Auto has always regarded "intelligent driving" as an important label. But in addition to the Zhiji LS6 APA automatic parking scraping I-shaped metal pillar. In January this year, a number of owners of Zhiji LS6 also reported that the vehicle had problems such as the disappearance of the car screen and the failure of the radar to alarm, which affected the normal use. Later, Liu Tao, co-CEO of Zhiji, also issued an apology statement on Weibo: "The main reason for the lack of screen display information and reversing prompt sound in some vehicles recently is due to a timestamp conflict problem in the PDC (reversing radar) prompt renderer integrated by IMOS, and the related problem has been fixed. "From the perspective of the industry, continuous optimization in intelligent driving will become an important challenge for Zhiji Automobile.
It is worth mentioning that Huaxi Securities said in a research report that intelligent driving has effectively influenced consumers' car purchase decisions, and in the future, the leading car companies that can identify differentiated selling points and leading technology will form differentiation in the era of "standard intelligent driving" and maintain a leading edge for more than two years.
At present, in the market, there are Wenjie, Xpeng Motors and AVATAR, etc., which are competing for the label of intelligent driving, and various car companies are launching urban NOA solutions. And with Tesla's FSD expected to enter the Chinese market, competition in the field of intelligent driving may intensify.
"This will really put pressure and challenges on the industry. Tesla's ability in vision is very strong, and it is intimidating in terms of many underlying technologies, underlying algorithms. But for Chinese smart electric vehicle brands, while there are challenges, there are also many opportunities. For example, we have a better understanding of the driving habits of Chinese people and China's road conditions. For Zhiji Auto, we can find a unique technological development path by accelerating iteration and data training to maintain differentiation. Liu Tao said in an interview with reporters.