In the era of AI, data has become a key decisive point for smart cars, which can train assisted driving to make smarter, human-like decisions, and promote assisted driving competition into the next track.
Musk once said: When it reaches 10 million Clips (training videos), the performance of the system becomes unbelievable.
Many brands have broken through this training volume, such as the three SUVs in this issue of “Super Unboxing”:
CarryingHuaweiQiankun ADS3.3.2 version of the assisted driving system Zhijie R7;
the first 10 million Clips version of the end-to-end assisted driving system Xiaomi YU7;
and Tesla Model Y, which is equipped with FSD and actively invites friends to compare.
So how many types of people are there in the performance of the high-training driver assistance system? Who is stronger between independent brand assisted driving and overseas FSD? How do the latest versions of Huawei assisted driving and Xiaomi assisted driving perform?
We tested these three cars in both urban navigation and active safety, including difficult scenarios that can widen the gap.
After watching the video, I believe you will have an answer to the above questions.
Since only the old Tesla can enable FSD, we use the old Tesla Model Y in the urban pilot link, and we use the new Model Y in the active safety testing link.
Urban pilot
The urban pilot test was conducted in Suzhou, with a total distance of more than 70km. It covers a variety of road conditions such as main roads in urban areas, narrow one-way roads with mixed flow of people and vehicles, and rural roads.
The passing ability of the intersection directly affects the correctness of the pilot route, so we have selected 7 representative intersections to take a look at the passing of three vehicles:
The most difficult intersection is that we need to turn right down the slope on the township road and pass through a bridge hole, there is no intersection sign, and the entrance cannot be found at a little distance, and the people in the car are a little unresponsive.
Xiaomi YU7 behaved as going straight and missing the intersection.
Tesla Model Y found the intersection but did not confirm it, and chose to continue going straight.
We deduce that the bridge hole building may have given Tesla the interference of obstacles on the right side of Tesla, which can also be found on Tesla’s official website.
For example, Tesla indicates that active cruise may not be able to grasp the distance when going uphill, so the slope road will affect Tesla’s judgment to a certain extent, and Tesla will try to find other exits when encountering obstacles such as road closures when driving normally, so this scene appeared.
The downhill road is a professional counterpart for the Zhijie R7, it did not hesitate, turned on the lights in advance, and found the right intersection to go down.
On the one hand, it is a hardware advantage, Zhijie R7 uses 192-line lidar, a detection distance of 250 meters, and has lidar point cloud data, which can model slope surfaces with low latency.
In addition, Huawei’s ADS solution has been optimized for slope roads, and the Zhijie R7 debuted intelligent downhill cruise last year, which can not only recognize the slope and automatically control the speed, but also adjust the suspension in combination with the Turing chassis to make the downhill more stable.
The scene of turning right on the dedicated lane shows the difference between Chinese and Western assisted driving.
The Zhijie R7 and Xiaomi YU7 easily turned right, but the Tesla Model Y did not adjust its strategy for non-motorized lanes, so it tried to squeeze into the narrow non-motorized lane after entering the right-turn lane.
It may be because I have not encountered such a problem in my overseas “hometown”, so I have not optimized it.
In terms of functional coverage, the Zhijie R7, Xiaomi YU7 and Tesla Model Y are very inclusive of different road conditions.
For example, narrow stone road streets and alleys, commercial food areas without lane lines, and remote villages can basically start functions.
Let’s take a look at the ability of the three cars to cope with various obstacles on the road.
The basic detour ability for pedestrians, non-motor vehicles, and temporary parking is not bad, but in the high-density obstacle area, the performance of the three is completely different.
The Tesla Model Y has a strong overall perception, but it seems to be a bit “nearsighted”.
For example, when it detoured the right melon basket in advance, it did not open up much distance because the left side was also limited, and when it was close, it almost pressed the melon frame.
It performs very well in the open area, but stagnates when close to people.
This is also a characteristic of Tesla FSD, and its approach is no problem in the United States, which has a good courtesy effect, and can usually guide people around to leave quickly.
However, there are many pedestrians shuttling and people and vehicles in China, which Tesla really can’t bear.
The performance of the Zhijie R7 is more mature, slowly circling left and right, moving through the crowd.
Xiaomi’s sensitivity is in the middle, with avoidance initiative, but when it encounters the other party coming faster, it will stagnate or require the driver to drive manually.
For the cone barrel occupying the road, the three cars can also change lanes to avoid it, which is more flexible.
Proactive security
When the pilot assist is not turned on, how much protection can the active safety of these three cars bring to the owner? We came to the professional closed test site and restored 5 high-frequency risk scenarios, and the test results are as shown in the figure:
Still water horse
The first four items have a starting speed of 35km/h, and the upper limit of the test is 125km/h.
The last front car cuts out the stationary rear of the vehicle because it mostly occurs at high speeds, so the starting speed is 80km/h, and the test limit is 120km/h.
The first stationary water horse project, Tesla made us lose 20,000 and 4,000 yuan!
Tesla’s AEB function does not respond to water horses, but its standard “pedestrian protection device” is very sensitive, and once it pops off, all relevant parts must be replaced, and maintenance costs are expensive.
Here, we speak for the majority of Tesla owners, and manufacturers optimize the “pedestrian protection device”, such as reducing maintenance costs and other operations.
Therefore, we replaced the new Model Y with a new Model Y in the follow-up AEB project, but limited by the “pedestrian protection device”, the testers are in the state of the limit as much as possible, predicting that the AEB function is not successful and avoiding people, and all data can be referred to as appropriate.
The AEB of the Xiaomi YU7 and Tesla Model Y did not respond to the stationary water horse.
The Zhijie R7 can still brake stably at a speed of 125km/h.
The Zhijie R7 will choose to trigger AEB later when it is in a low-speed state, but after speeding up, the trigger point of AEB will also be advanced, so its AEB performance is less likely to trigger by mistake, and at the same time, it is reasonable to determine the AEB trigger point based on its own speed and the necessary braking distance.
Adults cross the road during the day
Daytime adult crossing has been included in the latest national standard and is expected to be implemented in 2028.
Tesla has triggered braking at a speed of 35km/h but very late, which is similar to the point where the driver decides to manually avoid the limit.
Both the Xiaomi YU7 and the Zhijie R7 have achieved high speed obstacle avoidance, of which the Zhijie R7 once again brushed to 125km/h and braked.
During the day, non-motor vehicles pass diagonally
Non-motor vehicles cross diagonally during the day, restoring the traffic characteristics of some areas in China.
Tesla Model Y and Xiaomi YU7 have warnings but cannot brake to avoid obstacles.
The Zhijie R7 can warn of braking and avoiding obstacles at a speed of 105km/h, and will also warn and brake at a speed of 115km/h, but the braking force is not enough and will collide.
Backlit children’s ghost probes at night
The backlit children’s ghost probe at night is the most difficult project, the child rushes out from behind the obstacle car at a speed of 5km/h, the collision point is 50%, and the oncoming lights are extremely dazzling, and the driver finds it difficult.
In this environment, Tesla was unable to brake in time to avoid a collision.
Xiaomi YU7 and Zhijie R7 can brake and avoid obstacles at a speed of 55km/h and 65km/, respectively, and the overall performance is good, after all, in this low-visibility environment, everyone will generally drive slowly.
The front car cuts out and encounters the stationary rear of the car
The front car cuts out and meets the stationary rear of the car, leaving a short reaction time for the test car.
The Xiaomi YU7 can brake in the state of 80km/h, but it will be released after braking, so the owner remembers to make up the brake himself in case of this situation to avoid rear-end collisions.
When the speed is mentioned at 100km/h, the Xiaomi YU7 also reacts, but it can’t stop.
The Tesla Model Y can brake early warning at 80km/h, and the braking distance is more reassuring.
There is an early warning at 100km/h, but the braking point is too late and the driver chooses to avoid it manually.
The Zhijie R7 triggers AEB brake obstacle avoidance at speeds of 80km/h and 100km/h.
At a speed of 120km/h, the system predicts that the distance is not enough to brake, which directly triggers eAES enhanced automatic emergency steering assisted obstacle avoidance, that is, when the vehicle detects a collision risk and the collision cannot be avoided by braking alone, eAES will also control the vehicle’s steering for lateral avoidance while emergency braking.
We also tried a test speed of 130km/h once, which also triggered eAES safety avoidance.
The overall performance of the active safety part is that the Zhijie R7 maintains the consistent high speed and high level of Huawei’s system, and Xiaomi has also made great progress after upgrading to the new version, while Tesla has not yet made efforts in the AEB function.
Written at the end
Big data has driven the rapid increase in the availability of urban pilot assistance, and the system has been able to cope with most road sections to help drivers reduce fatigue.
However, the risk is usually in an instant, sometimes pilot assistance or AEB can be dealt with, bringing the safety redundancy of “human-machine double insurance”, but there are definitely situations that cannot be dealt with, the driver should concentrate on driving and make reasonable use of the value of the assisted driving system.
So after watching this video, who do you think is the most satisfied with the performance of Zhijie R7, Xiaomi YU7, or Tesla Model Y?
Has the performance of upgrading the latest version of Zhijie R7 and Xiaomi YU7 exceeded your expectations? The comment area is waiting for your answer.
