Automatic assisted driving re-exposed safety accidents, Xiaopeng Motors and the owner of the car fell into Rashomon.
A few days ago, Mr. Deng, the owner of Xiaopeng Automobile, released a video to defend his rights, saying that after turning on automatic assisted driving on the national highway for more than ten kilometers, a rollover vehicle appeared on the road ahead, and the P7 he drove did not have any alarm and deceleration, and directly hit it. There was an emergency braking after the abnormality was found, but the car did not react at all. It is worth noting that according to the driving recorder provided by the owner, the acceleration value before the impact has risen.
The acceleration value rose before the impact, and some netizens questioned the owner of the car for the brake as the accelerator, but the owner said that he had 7 years of driving experience.
As of now, Xiaopeng Automobile has not yet published the owner's operation record, whether it is a brake failure or the owner's misoperation, there is no conclusive conclusion.
Just last month, Tesla had a similar accident. Ms. Lin, the owner of the Model3 who has been carrying the car for less than two months, said that Tesla's kinetic energy recovery function should be to release the throttle and slow down until it stops, but I did not step on the accelerator, the speed did not decrease, and even had a tendency to speed up. "Later, I slammed on the brakes many times, but the car was not controlled, and in the case of not daring to judge the condition of the car on the right side, it directly hit the wall, and the airbag popped out."
According to the analysis data provided by Tesla, Ms. Lin last pressed the brake pedal at a speed of 30.04km/h, and after pressing the brake pedal for 0.2 seconds, the brake master cylinder pressure reached 138.9 bar. 0.1 seconds before the collision, Ms. Lin also stepped on the brakes. But before the car crashed, the braking system did not get out of control.
Tesla also refused to provide complete driving data, proposing to tow the accident car to tesla stores for testing. However, Tesla does not have judicial qualifications by default, and only provides report results from the perspective of vehicle maintenance.
For the cause of the safety accident, Xiaopeng Automobile responded to the Observer Network, "Regarding the accident reflected by the owner Mr. Deng that he collided with the accident vehicle parked in front of him when using ACC+LCC (Adaptive Cruise Control & Lane Centering Function), our company attaches great importance to it. He deeply regretted the accident of the vehicle and expressed concern about Mr. Deng's physical condition.
"After investigation, on March 13, the owner of the car, Mr. Deng, drove the Xiaopeng P7 in the local highway section and collided with the overturned vehicle parked in front of him, and the accident did not cause injury. After the accident, through the background collision monitoring, our customer service staff called the customer for the first time to inquire about the situation and give assistance and guidance. The preliminary judgment is that the owner did not maintain the observation of the environment in front of the vehicle and take over the vehicle in time during the process of making the ACC + LCC (Adaptive Cruise & Lane CenterIng Function). After the investigation and judgment of the local traffic police department, the conclusion is also in line with the above situation.
"Here, we need to remind the majority of drivers once again that in the process of using the assisted driving system, it is necessary to maintain observation of the surrounding environment of the vehicle and follow the relevant content of the safety guidance test before the opening of the driving assistance function to ensure driving safety."
"Customer experience is the top priority of Xiaopeng Motors, and after the accident, we have provided Mr. Deng with a free scooter for the first time. We will also actively communicate with Mr. Deng about follow-up maintenance matters to help car owners resume use as soon as possible. ”
Judging from Xiaopeng Automobile's response, Xiaopeng Automobile believes that the accident was caused by the improper operation of the owner and does not admit that there is a problem with its auxiliary driving system.
According to the data, Xiaopeng's driver assistance XPILOT includes functions such as Lane Center Assist (LCC), Adaptive Cruise (ACC) and Automatic Lane Change Assist (ALC) functions. Among them, LCC mainly detects the lane line on the road through the camera on the windshield, and assists in controlling the steering wheel, so that the vehicle remains in the middle of the two lane lines. ACC is perceived by radar sensors and cameras, detects the vehicle in front and the current speed limit, and assists the owner in controlling the acceleration and deceleration of the vehicle. When it detects that there is no car in front, the system maintains the set speed to automatically move forward; when it detects that there is a car in front, it maintains the distance, automatically follows and slows down.
If the owner of the Xiaopeng car uses the assisted driving system for the first time, the operation interface of Xiaopeng Automobile has a mandatory test answering session, which marks the unrecognized use scenarios of the ACC+LCC system. The answer shows: low-speed or stationary engineering vehicles /sweepers; accident cars that stop horizontally; vehicles with congested or waiting for red lights; large rocks; tripods; protruding isolation belts, water horses or cement piers, etc., such objects xiaopeng automobile ACC+LCC system can not be recognized, requiring the owner to take over the vehicle immediately.
It can be seen that the objects appearing in the above scene are all stationary objects. At this stage, whether it is the pure visual perception dominated by cameras that relies on pixel offset judgment or the point cloud model composed of "lidar + high-precision map", there has always been a blind spot in the perception and recognition of stationary objects or slow-moving objects by vehicles. Static objects require machine learning to train to recognize objects, but static objects have many categories and shapes, requiring a large number of sample training. If the perception system misjudges the road scene, other systems, including the AEB system (Automatic Emergency Braking System), will not react.
Ideal CEO Li Bin once said, "At present, the combination of camera + millimeter-wave radar is like a frog's eye, which is good for dynamic objects and almost incompetent for non-standard static objects." The progress of vision at this level has almost stagnated, even if it is dynamic, the recognition rate outside the vehicle is less than 80%, and it should not be used as automatic driving. ”
In addition, the identification of objects by on-board cameras is also affected by various factors such as weather and lighting conditions, and the speed of the vehicle.
Tesla CEO Musk once advertised in 2020 to achieve full self-driving, and two years later, Musk also reneged on his promise, acknowledging that "fully autonomous driving is a problem because it needs to solve most of the artificial intelligence problems in the real world."
According to the global level 5 classification method for autonomous driving, most of the assisted driving functions are still at L2 level, or some manufacturers have advertised that they have reached L2.9. In other words, the L3 level is still not reached. Among them, L2 level clearly stipulates that when using the function of assisted driving, the driver needs to observe the surrounding environment and be ready to take over at any time.
Although new forces such as Xiaopeng, Ideal, and Weilai have repeatedly emphasized in public that "assisted driving" is not "automatic driving", there are still car owners who lack a clear understanding of the assisted driving system, resulting in some safety accidents.
This article is an exclusive manuscript of the Observer Network and may not be reproduced without authorization.