In the context of the increasingly crowded new energy vehicle track, auto companies have begun to spell speed in addition to research and development and marketing.
Recently, Tesla's 12,000-acre Plant in Texas, USA, began production. This means that Tesla will have a larger site to achieve vertical integration of large components, and all aspects of car manufacturing can be completed in one factory.
According to Tesla's design plan, the production line of the Texas factory can produce a body every 45 seconds, and a Model Y can be built every 10 hours, and at this rate, the maximum production capacity will reach 1 million vehicles per year.
Looking at China, following the new forces of car manufacturing, the addition of baidu, Huawei, millet and other technology companies has also added a fire to new energy vehicles.
In March last year, Baidu and Geely established Jidu Automobile, Baidu's autonomous driving technology and Geely's mature vehicle manufacturing experience have attracted much attention since its inception.
Where to go on the shoulders of giants has become a widely concerned issue in the industry.
The answer given by Jidu is the "weekly change" speed, which makes up for the disadvantage of late establishment time with efficiency. In line with the research and development speed of the set, it can be seen that almost every two months the set will actively release the internal dynamics to the outside world.
Concentration speed has become a new phenomenon of car building.
When the set degree was 100 days, the set degree published a page of timetable, which marked various core time points, such as the company registration in March, the end of the modeling review in July, the opening of the robbery contest in the automobile industry, the signing of the contract with Hesai Technology in August, and the scale of 1,000 people in October.
The time came to April 18 this year, more than a year after the establishment of Jidu held an online media communication meeting, Jidu intelligent driving head Wang Weibao, chatted with the media about the new progress of Jidu.
01
SIMUCar enters 2.0 and dual lidar is on board
Yesterday was originally the day when Jidu held Robo Day and released The first concept model of Jidu, but due to the impact of the epidemic, Jidu Robo Day was postponed to June, and the activity day was changed to a media communication meeting, and Jidu released two highlights with more restraint.
Fast progress: JIDU SIMUCar has entered the 2.0 stage, JET, Nvidia chips, lidar on the car test;
True redundancy: Equipped with dual lidar, Jidu will develop two sets of pure vision and lidar independent of each other.
Starting from SIMUCar, the full name of SIMUCar is software integrated analog prototype car, which can be used to pre-position the research and development of intelligent cockpit and intelligent driving by decoupling hardware and software.
To put it simply, originally, car companies generated prototype cars after the underlying hardware was complete, driving on the live road, so as to achieve intelligent driving data accumulation and live testing, and now in order to save time, the live test is transferred from offline to online to achieve a parallel development mode of software and hardware.
SIMUCar is not a fixed model, it will continue to advance with the progress of the development of the set, and eventually form the "automotive robot" with self-learning and evolutionary ability expected by the set degree.
In this process, Jidu Intelligent Driving Wang Weibao divides it into three stages:
SIMUCar 1.0 is the basic capacity building for the primary school stage, mainly to achieve intelligent assisted driving;
SIMUCar 2.0 is the secondary school stage, in which the domain controller gets on the bus and evolves to the state of mass production of automatic driving;
SIMUCar3.0 is the university stage, the set degree will enter the mass production verification link, the software system and the vehicle formal joint commissioning, in the debugging process to improve the special ability of the intelligent driving system, according to the owner's driving habits to learn and upgrade, and finally achieve self-learning, self-learning ability.
Wang Weibao introduced to the heart of the car that once SIMUCar completely runs through the above three stages, it will achieve the goal of software-defined hardware, achieve rapid iteration, and shorten the production cycle of smart cars.
At present, SIMUCar is in the second phase, Wang Weibao revealed: Nvidia Orin and Hesai AT128 semi-solid-state lidar have been on the car to participate in the SIMUCar 2.0 test.
Although Jidu did not disclose the number of NVIDIA chips, it confirmed that it would be equipped with dual lidar to participate in the test and gradually approach the mass production state.
02
Pure vision and lidar are independent of each other
Creating "true redundancy"?
The hardware ambitions displayed by Jidu in dual lidar and NVIDIA chips also point in one direction - to improve "security capabilities". To this end, in addition to the powerful hardware, Jidu has also prepared a "true redundancy" solution of pure vision + dual lidar.
According to the heart of the car, in the early SIMUCar 1.0, Jidu only relied on pure vision to achieve dual-domain integration of urban roads and high speeds.
At present, Jidu is testing two automatic driving schemes of pure vision + two Hesai lidar to achieve independent operation of the two sets of schemes and ensure the safety of traffic participants inside and outside the car.
Among them, the most noteworthy is that these two Hesai lidars are on the car. The HESAI AT128 integrates 128 independent lasers and achieves a global resolution of 1200x128 and a point cloud count of 1.53 million per second by simultaneously firing 128 lasers into three-dimensional space.
Combined with these performances, the performance of the AT128 can be regarded as the "first echelon" of the domestic market mass production vehicle specification-grade lidar. At present, in addition to the set, AT128 lidar is also installed in the ideal L9, Lotus Electre, Gaohe HiPhi Z and other models.
It is undeniable that the two lidars do bring "double feeding" to the concentration.
In the data disclosed by Jidu, dual lidar can double the point cloud, making the identification accuracy higher and safer, and at the same time, it can be redundant with each other and provide more reliable security.
Are two strong enough?
Jidu released a test map of Jidu SIMUCar2.0 driving at night with dual lidar to the outside world:
Narrow roads will be cars
Turn left unprotected
There is protection to turn left
Throughout the press conference, we also noticed that Jidu referred to pure vision + lidar solutions as "truly redundant" perception systems.
According to Wang Weibao, this truly redundant system of Jidu, supported by the bottom Baidu ApolloL level 4 automatic driving capability, can achieve a highly self-driving, efficient and collaborative true redundant high-order automatic driving scheme in the dual-perception system.
Here it must be emphasized that the so-called redundancy is not the integration of two systems, but the ability to achieve mutual independence and mutual backup under the same conditions, which plays a role in improving the safety of automatic driving.
Similar to Jidu, at about the same time, Intel's self-driving company Mobileye also proposed a "true redundancy" scheme.
Mobileye CEO Amnon also introduced that his environmental sensing solution contains two independent subsystems, one using only cameras, the other using radar | lidar, and the two subsystems are backed up from each other.
On April 13, Mobileye's self-driving taxis with "true redundancy" self-driving systems were field-tested in Israel. In the 40-minute video of the actual test, Mobileye's test vehicle safely and smoothly passed the driving scene of poor lighting and complicated road signs.
Mobileye self-driving taxi test footage
Although the industry is still thinking and arguing about the "true redundancy" scheme, some people believe that this increases the cost of building a car and is not too necessary.
But obviously, There are more convincing reasons why Jidu and Mobileye chose the "true redundancy" scheme, such as "true redundancy" can greatly reduce the amount of data required to verify the perception system, and secondly, when one of the independent systems fails, the vehicle can still continue to drive safely.
03
Good wind with strength, set degree "leap of technology"
The so-called tube peeping leopard, the performance highlights released by the set degree are only one aspect.
On the other hand, since Jidu is a joint venture between Baidu and Geely, Jidu is naturally also named "Baidu Car".
Then, in addition to asking the reasons for the "real redundancy" solution of jidu equipped with shuanghe sai lidar in terms of performance, it is more necessary for us to explore the capital phenomenon after the performance highlights of jidu.
In fact, whether it is lidar or "true redundancy" solutions, there is a similar figure behind it, that is, Baidu.
Hesai Technology, which provides AT128 lidar for Jidu, has always been closely related to Baidu.
In 2017, Hesai and Baidu Apollo jointly released a set of autonomous driving developer kit Pandora, and then participated in the research and development of Apollo Robotaxi.
According to relevant reports, Baidu holds 7.88% of Hesai Technology and is the second largest institutional investor of Hesai.
Therefore, it can be seen from the investment structure that Jidu chose Hesai, which is held by Baidu, as a lidar supplier, with the intention of using stable capital relations to hedge with the current unstable supply chain, and to provide guarantee for Jidu to carry Hesai AT128 mass production on the car.
In addition, Baidu is also the basis for Jidu to choose pure vision + lidar solutions.
In fact, there has always been a dispute between "pure vision solutions" and "lidar" routes in the field of autonomous driving.
The former takes machine vision as the core, mainly uses cameras to achieve automatic driving, in line with the logic of the human eye, the cost is low, and it is currently mainly Tesla FSD that adopts this program.
Although the lidar scheme can make the perception more accurate and accurately restore the road condition information, the cost is still relatively high, basically at the price of $1,000.
In order to make up for the shortcomings of these two technical routes, a third route solution combining pure vision combined with lidar has gradually emerged in the industry and academia. Baidu autonomous driving is a small number of domestic pure vision, lidar two routes have a deep accumulation of enterprises.
Wang Weibao said bluntly at the media briefing: "Since Day One, we have worked closely with Baidu Apollo, and the solutions of the entire automotive robot still mostly use Baidu's ability to atomize based on Apollo Robotaxi."
In our understanding, the so-called atomization capability is based on Baidu Apollo Lite's ability to achieve L4 autonomous driving in the urban area based on pure visual perception alone + Robotaxi's 25 million kilometers 0 accident road test data accumulation advantage.
Therefore, after SIMUCar1.0 achieves pure visual automatic driving, the addition of lidar is the icing on the cake, providing algorithm and hardware safety redundancy capabilities for the final set of "true redundancy".
Good wind with strength, send me to Qingyun, for Jidu, Baidu is the "East Wind", Jidu can achieve acceleration under the epidemic is also expected.
Regarding more dynamics of Jidu's car manufacturing, Jidu said that it will be left until robo Day at the end of June to release fierce materials, when Jidu's new car will be officially unveiled. However, such a tight spread rhythm of the set degree also released a positive signal of "all in control" to the outside world, which is not easy in the epidemic environment.
As we all know, affected by the epidemic, the yangtze river delta region of the automobile supply chain is broken, resulting in car companies into a production crisis Huawei executive director Yu Chengdong, Xiaopeng Automobile founder He Xiaopeng have expressed concerns in the circle of friends, He Xiaopeng is simply and clearly pointed out that "if Shanghai and the surrounding industrial chain enterprises can not find a way to dynamically resume work and production, in May all China's automakers may have to stop work and production."
Because Jidu is in the research and development stage, coupled with the protection of Baidu technology and supply chain, the crisis of "shutdown and production stoppage" has not affected Jidu on a large scale.
Zhang Jiahao, head of marketing growth of Jidu, also said that at present, Jidu has not been affected by the epidemic, and there is a special supply chain supply team inside to ensure the price, quality and supply of parts, etc., to ensure the launch of mass production vehicles in 2023 as originally planned.
What exactly does this production car look like? Jidu sold a pass, and only released a "side face photo" of Jidu's car.
According to the photos, the dual lidar is arranged on the left and right sides of the front cabin cover, which is very different from the ideal and most models of Weilai to arrange the lidar to the roof position.
After the photo was released, it attracted the onlookers of Ideal Auto CEO Li Xiang, who thought that the roof layout of lidar was the optimal solution, while Xia Yiping, CEO of Jidu Automobile, believed that the problems solved by lidar in the roof layout and the layout on both sides, product capabilities and safety were different.
So, which solution are you more optimistic about? Welcome to the comments section.
"The Heart of the Car · The connoisseur said" trailer
On Thursday, April 21, 20:00-21:00, Box Cars founder and president Zhang Shuguang will be a guest car heart · Experts said, talk about "travel market nuggets: where is the opportunity point of the B-end track of new energy vehicles?"