On April 18, Jidu Automobile announced the latest progress of its first model, which can be seen from the picture with a through-line light strip, ADAS working indicators on the wheel headers, and more conspicuously, the two lidars standing on the cabin canopy.
At this communication meeting, Jidu Automobile also confirmed that the model of the lidar is Hesai AT128, which adopts a telescopic design.
The layout method has just been exposed, and Li Xiang, the founder of Ideal Car, shared his views under the Weibo of The 42nd Garage:
"Putting one on the roof and two on the cover or bumper makes no difference in performance, and even a single one on the top of the head will perform better." However, in terms of pedestrian collisions, repair costs, and vibration control (linking the main body), the roof is optimal. The only problem is that the shape of the roof lidar will need to be adapted, because it is too much like a mobile suit - Gundam."
Shortly after, Quarterly Auto CEO Xia Yiping replied to the comment, saying:
"There is still a difference, 120 degrees OF FOV and 180 degrees of FOV are still not the same, and the solution of the corner case is not the same, so there is still a difference in the ability and experience of the product and even the security."
Standing at the time point of the large-scale lidar on the car, we have seen a number of car companies give "overhead", "cover", "bumper", a variety of layout schemes, so take this opportunity to do a wave of inventory for everyone, see how each car company is choosing, what does it mean behind the lidar layout?
01
Let's take a look at the scheme of set size
As can be seen from the figure, the 2 lidars of Jidu are located at the left and right ends above the locomotive, which is the first layout of Jidu, and adopts a telescopic design.
Two radars complete the detection coverage of 180° at the front of the vehicle, while the core area of the front of the vehicle also achieves 60° dual radar coverage.
From the official website document of Jidu, we have summarized 3 major advantages:
The FOV of the front of the car with a dual radar level of 180° covers the driving scenes such as "ghost probes" and left and right obstructions, and has a stronger ability to recognize pedestrians or obstacles traversing left and right;
In the area of 60° FOV directly in front of the vehicle, the dual lidar can be doubled and overlapped, and the point cloud data on the target is more, and the recognition accuracy is higher;
Dual lidar can be safely redundant with each other, which is more reliable than a single lidar solution.
First of all, compared with 120°, the detection angle of 180° reduces the visual blind spot of 30° on the left and right sides, and the reduction of the physical blind zone will inevitably improve the detection ability of ghost probes and other scenes.
Moreover, the radar covers both the side and the front of the car, and the entire entry trajectory of the jammed vehicle can be monitored by one radar.
Here is a more typical case of ghost probes, the vehicle on the right side of the road, obscuring the detection of the radar on the right, and the radar on the left side can just see the area in front of the vehicle due to the more external position, or has a better field of view.
Just like our human eye, it blocks the right eye, but thanks to the large viewing angle of the left eye, you can also see the picture of part of the blind spot of the right eye.
The camera and the right Lidar did not detect pedestrians in the blind spot, the left Lidar saw
The second advantage is a dual radar perception coverage of 60° in the middle area, which allows for denser point cloud information over just one lidar, resulting in more accurate perception results.
However, the premise for obtaining more accurate perception results is that the results of the two radar outputs must be highly compatible with the fusion algorithm of lidar data.
The third point is redundancy, where redundancy has 2 layers of meaning, jidu official said: "Dual lidar can be each other's safety redundancy, which is more reliable than the single lidar solution."
There is no doubt about this, when one radar fails, the other radar can continue to work, playing a role in a certain degree of redundancy, just like the two motors of a dual-electric model are redundant with each other.
The second layer of redundancy is at the system level, Jidu official said: "SiMUCar 2.0 stage is testing the development of pure vision and lidar autonomous driving solutions, is a separate dual system, can achieve 'true redundancy' after mass production. Compared with the traditional solution, the two sets of automatic driving solutions of Jidu are mutual backup and complement each other, and the two sets of solutions can be both efficient self-driving and efficient collaboration."
So at the system level, vision and lidar are two systems that can work together or independently.
Here I also carefully guess that Jidu will be like Mobileye, after the visual system failure, the radar system can continue to ensure the normal operation of the vehicle, in other words, relying on the visual set of auxiliary driving can work independently, relying on lidar can also work independently.
However, this view has also been opposed by industry leaders, and Su Jing, the former head of Huawei ADS, once said in a media interview: "There is a head autonomous driving company that proposes true redundancy, taking radar and LIDAR as a subsystem, taking pure vision as a subsystem, independent testing, and multiplying the takeover rate of the two subsystems to achieve a statistical reduction in the required test mileage." A question to answer:
"Frankly, I guess it was written by their Marketing, definitely not by their developer, otherwise I would have doubted their ability to do it." What really determines your takeover rate is not only your perceptual system, but also has a great relationship with your regulation, even greater than the perceptual system. These systems are not designed with what you call true redundancy, right?
Second, the vast majority of cases that are difficult to handle, you can't handle with 80 times the sensor. I bet you. So this kind of multiplication algorithm to do statistical logic is very absurd.
True redundancy is a very Marketing statement, you have to do a good job of perception, you should do sensor fusion, not do redundancy, do redundancy is a very serious waste of sensors. Their level of technology is definitely not like that."
So for the concept of the second layer of redundancy, we take a wait-and-see view.
Overall, the effect of these 2 radars is not bad, but Li Xiang expressed some different views.
02
The public says that the public is justified, and the mother says that the mother is reasonable
Li wanted to say: "Putting one on the roof of the car, and putting two on the cover or bumper, there is no difference in performance, and even the performance of a single one on the top of the head will be better." However, in terms of pedestrian collisions, repair costs, and vibration control (linking the main body), the roof is optimal. The only problem is that the shape of the roof lidar will need to be adapted, because it is too much like a mobile suit - Gundam."
Although this sentence is not long, it reveals Li Xiang's consideration of 6 dimensions when deciding the position of lidar:
Probing performance
Number of radars
Pedestrian Collision (Safety)
Repair costs
Vibration control
Exterior styling
The core purpose of lidar is to achieve higher-order assisted driving, so performance must be the first element.
The biggest advantage of placing the radar on the roof of the car is that the detection effect is excellent, just like the higher the sentry building, the wider the detection range of radiation, and the same is true of lidar.
Therefore, from the functional point of view, within a reasonable range, the higher the position of the radar, the better the detection effect, so placing the radar on the roof of the car must be the optimal solution. Correspondingly, second on the cover and third in the bumper.
However, one of the core points that Li xiang mentioned here is: "Putting one on the roof and putting two on the cover or bumper makes no difference in performance, and even the performance of a single one on the top of the head will be better."
That is to say, in the ideal L9 and set degree of the two layouts, height is more important than quantity.
Coincidentally, the ideal L9 lidar is also the AT128 provided by Hesai Technology, so we can exclude the impact of radar performance and directly discuss the impact of location and quantity on the detection effect.
Subsequently, Xia Yiping, CEO of Quarterly Automobile, replied: "There is still a difference, the 120-degree FOV and the 180-degree FOV are still different, and the corner case solved is also different, so there is still a difference in the ability and experience of the product and even the safety."
The central idea of this statement is that a wider detection angle can lead to greater capabilities, solving more corner cases, and better security.
Obviously, the two big guys are not talking about a point, Li wants to say: I am high in height, 1 top you 2; Xia Yiping said that I 2 to see widely, extreme scenes and safety performance is better. But it is clear that both sides want to highlight the goodness of their products through the advantages of their own.
However, it is clear that higher altitudes will definitely have better detection results, and as for the size of the detection FOV, it depends more on how the car company balances the relationship between the additional 60° detection angle, the realization of safety redundancy and the cost of twice the cost of lidar.
Another undeniable problem point is that placing lidar on the roof of the car has lower maintenance costs, better pedestrian safety collisions and better mechanical stability.
At present, a large number of models on the market put lidar in the position of the middle network or bumper, the Audi A8 that produced lidar for the first time, the new Audi A7L, Mercedes-Benz S, all continued this design, and the new forces of Xiaopeng, Avita, Jihu, Salon and other models also adopted such a design.
But the bumper was designed to protect the body, better flexibility and cheaper cost, which can reduce maintenance costs in the event of a minor accident.
Putting expensive lidar into the bumper is obviously contrary to the original intention of reducing maintenance costs, so it is doomed to have extremely high maintenance costs when such models rear-end.
The design of placing radar on the roof of the car avoids this problem to a large extent, after all, when the intensity of an accident is enough to cause the roof to deform, the car may not be far from scrapping.
The second issue is pedestrian safety, with the increasingly stringent car collision standards, the requirements for pedestrian protection in collision safety are also increasing, and vehicles need to be fully pedestrian collision safety in the design stage, which also poses a higher challenge to the layout of bumpers and radars.
In addition, as a high-precision perception component, lidar is not picky about the environment, and the use environment of the car is destined to be harsh and harsh, so the radar is elevated, and the impact of sediment and vibration on the radar can also be reduced to a certain extent.
03
Lidar putting on the roof is all an advantage?
Of course not.
The biggest problem on the roof is "user acceptance", after all, the emergence of lidar has broken the traditional styling design of vehicles, and at first glance, it is natural to feel "awkward" or "ugly".
The peak of the controversy was when NIO had just released the ET7, the first design to put lidar on the roof of a production car. Some users can accept the horns of the roof, but there is a larger wave of users who cannot.
In Wei's view, the lidar standing on the roof of the car will become a symbol of "high-end assisted driving" status, just as the rear wing is a symbol of the status of performance cars.
Also placed on the roof of the car are ideal, Feifan, GAC, Weima and other models, seeing the shape of these cars, it can be said that Weilai has made great efforts in the design, doing its best to integrate the lidar into the body and reducing the sense of abruptness.
Models that place lidar in the position of the bumper can greatly reduce the presence of lidar.
At the same time, putting lidar on the roof of the car will also increase the wind resistance coefficient of the whole vehicle, bringing greater wind noise, and has higher requirements for the NVH of the vehicle.
04
Write at the end
Where lidar is placed depends on how the car company balances the relationship between "perception effect", "appearance and aesthetics", "cost", "collision safety and maintenance cost".
If you want to maximize the use of lidar, it will generally be placed on the roof, but the "harmonious" appearance is abandoned behind it, and like the Xiaopeng G9, the dependence of the automatic driving team on radar is not so high, and it is more of a role in filling blindness, so the priority of appearance can be put in a higher position.
And the set degree can be said to take the advantages of the above two respectively, with a good detection ability, but also to ensure the harmony of the appearance, of course, now the next conclusion may be too early, and in other words, the set degree of the program also has 2 common shortcomings.
Finally, having lidar does not mean having a strong assisted driving, the advantages and disadvantages of the position do not represent the strength of the ability, lidar is just a hardware, the actual experience depends on the more core software algorithms.
Author: June 3
Edit: June 3
These are also worth reading