For this, the True Scenario Intelligent Assessment intelligent safety chapter will focus on 10 traffic accident scenarios caused by high-frequency, sudden events and adverse weather conditions. The test conditions will cover both daytime and nighttime, and special invitations have been extended to the currently hottest "six heroes" - Xiaji M9 (parameters|inquiry), Xiaomi S7, Tesla Model 3, Jixue 007, Weilai ET5, and Xiaopeng P7. See if it's safe online or just a show; can you really do it, or just make a fuss? This article will reveal the truth one by one.

Hardware configurations have their own strengths and weaknesses, with no significant price difference.

Intelligent safety-related hardware configuration, the sixth car positioning highest question boundary M9 is the only one equipped with 192-line laser radar, while Xiaopeng P7 is the only one installed with two laser radars, and in terms of millimeter wave radar and camera numbers, Xiaopeng also belongs to the sixth car's thick. Compared with it, the latest released Xiaomi SU7 did not show a late-comer advantage, a 128-line laser radar from Weicai Technology was placed on top of the car, three millimeter wave radars and a total of 11 cameras "armament" could only be considered normal performance. Of course, the most biased one in the whole field is still Tesla Model3, the pure visual scheme means that no radar is equipped, and the quality of the dual 5 million pixel front-facing camera also does not stand out.

Assisting driving chip, except for the Me9 without official announcement, Model3 insisted on using self-researched HW 4.0 chip, while the other four cars chose to purchase from NVIDIA, especially the NIO ET5's "ADAM Central Computing Platform", even used 4 NVIDIA Orin X chips, finally reaching a powerful 1016 TOPS. It seems that the price difference among the six cars has not been directly reflected in their intelligent hardware levels, only actual testing can see who is strong and who is weak.

For this, on the test equipment, we not only used DRI 3D mock car, engineering car, tricycle, motorcycle and full set of mannequin target objects, but also relied on RT, VBOX and i-TESTER etc. advanced positioning devices to accurately analyze the actual speed of the test vehicle, collision time, how far away from the collision point did it start braking? What was the maximum deceleration when braking? How far away from the collision point was it after braking, etc. In terms of vehicle settings, we will upgrade all vehicle systems to the latest and open ACC, AEB, etc. active safety functions as needed, while setting uniform sensitivity and following distance in the middle gear, night subjects will turn on automatic near and far light function.

Epidemic event

First, let's look at the four subjects that are custom-made for daily high-occurrence events: against oncoming cars invading reverse driving, intersections electric motorcycles turning left to cut in, multiple pedestrians crossing and children stopping suddenly in emergency. They have a commonality: all of them are commonly seen on daytime urban main roads, and it is highly probable that they were caused by the other party's traffic violations first, resulting in the sudden change of movement state of the traffic participants, which has put a lot of pressure on the ACC and AEB systems of six vehicles.

For example, when a vehicle invades the opposite lane while reversing, the main simulation is just the scenario of vehicles invading the normal driving lane in urban narrow roads and rural roads passing through lanes. The test was conducted separately under the conditions of ACC-related functions being turned on and AEB working, with the test speed being 60km/h and 80km/h respectively, and encountering a fake car traveling at 30km/h head-on. Whether adaptive cruise control is turned on or pure line-following AEB testing, it can stabilize braking and stop at both test speeds of 60 and 80 km/h. The image warning and alarm sound are also quite in place. At the same time, the braking intervention was prompt and had a lead time. Even after complete braking, the distance to the fake car could be stably maintained at around 5-7 meters, which is quite nice. The only thing that's not perfect is that when the opposite vehicle is clearly approaching, the rear end of the vehicle may flash on the central control screen.

Compared to the legendary Silicon Valley, the Weilai M9 and Xia Li ET5's composure and calmness can only rely on ACC's protection. The warning prompts are orderly, the braking intervention is stable and gentle, which is still our favorite style of "braking" experience. Especially for the Weilai M9, there were dodging actions within the lane lines near the braking point. In the AEB test scenario, both cars were completely lost, with a speed of 60km/h, Weilai did not have any effective warning of impending danger, only displaying an emergency braking prompt image and alarm sound just before impact, without any braking intervention throughout. The Xia Li ET5, although triggered forward collision warnings earlier, also lacked braking intervention. The "Three 7 Combination" - XPeng P7, Xiaomi SU7, and Geely 007 - either had braking but with insufficient deceleration force, or released the brake just before stopping, or only triggered a collision warning without braking at all. In terms of data, Xia Li ET5's relative distance to braking was significantly earlier than Weilai and Geely, and its maximum deceleration during braking reached 1.18g.

The oncoming car in a narrow road was caught off guard, and the two-wheeled vehicle turning at an intersection was even more difficult to defend against. Let's take a look at how each of you performed in intercepting the left-turning two-wheeled vehicle under AEB conditions, with test speeds of 20, 40, and 60 km/h. The six cars were divided into three groups: full pass, partial pass, and complete failure to pass. The Tesla Model 3, Xiaopeng P7, and Xiaomi SU7 all failed the basic 20km/h test, but the Tesla Model 3 and Xiaopeng P7 had a near-collision experience with one-foot braking and triggering an alarm. Only the Xiaomi SU7 completely ignored the pedestrian in its sights, riding recklessly throughout without any regard for safety.

While the NIO ET5 and the Weilai M9 have been calibrated for such conditions, they can successfully avoid accidents at 60km/h. In contrast, Xpeng's success stopped at 40km/h, and while it had a near-miss at 60km/h, with interior ambient lights and exterior hazard warning lights, the braking intervention was too late to avoid an accident. Data shows that NIO ET5 brakes earlier than Weilai and Xpeng, and its maximum deceleration during braking reached 1.18g.

Again, look at the pedestrians crossing the road at a variable speed. It mainly simulates the normal straight-line vehicles at an intersection, suddenly encountering two adults and one child violating the law to cross the road, and the child suddenly emerging in an emergency situation. The test is performed under AEB conditions, with vehicle speeds of 30km/h and 50km/h, and a child dummy speed of 5km/h, let's see how it performs.

Of course, the sensitivity of intelligent safety systems also has its limits. A strong false alarm can lead to a new accident. Let's look at the child scenario where the child stops in an emergency. It mainly simulates a normal straight run through an intersection, with a child pedestrian violating the rules and stopping before crossing the road again. The evaluation standard is very simple: there must be a warning, but strong braking that increases the risk of a rear-end collision cannot be allowed. The test is also performed under AEB conditions.