The “social experiment” of technology giants’ cross-border car manufacturing is one step closer to the result.
Backed by Baidu and Geely’s Chidu Automobile, it has whetted people’s “appetite” since its establishment: whether it is Baidu founder Li Yanhong who said that “Chidu’s self-driving technology will lead the Tesla generation”, or Chidu CEO Xia Yiping announced the “2880” strategy: by 2028, it will have the ability to deliver 800,000 vehicles for the whole year.
From its establishment in March 2021 to the present, Jidu has completed the rapid expansion of the team, product definition, and concept car debut in the past more than 600 days.
Today, Jidu is one step closer to the mass production of its first car. Recently, Jidu’s first car robot ROBO-01 moon exploration limited edition was officially released. The new car is priced at 399,800 yuan, which is 2,000 yuan more expensive than the Model Y top model. This is a joint model of the lunar exploration project launched by Jidu and China Lunar Exploration. It is upgraded from the Jidu ROBO-01 concept car, and at the same time adds some elements of “Chinese lunar exploration”.
It is understood that users of the limited edition of the lunar exploration will receive a series of exclusive rights and interests of high technology, among which the rights of “lifetime high-end intelligent driving package” are the most dazzling. According to the official introduction, this life-long intelligent driving right is “following people but not cars”, and the rights of users of the limited edition of the moon exploration will continue when they replace any integrated models. At the same time, these users also have the opportunity to observe the launch of the Chinese space mission rocket on site.
The limited edition of ROBO-01 Moon Probe has only 2,000 seats. The first batch of 1,000 Chief Experience Officer seats have been reserved. At the press conference, Xia Yiping announced an additional second batch of 1,000 seats. At the press conference, Liu Cixin, the author of the science fiction novel “Three-Body Problem”, appeared in the form of a video as the owner of the limited edition 004 lunar exploration.
As a car that benchmarks Tesla Model Y, although Jidu ROBO-01 has not announced a detailed configuration table this time, it can still be seen from the details of its unique thinking and cognition in smart electric vehicles.
The “ambition” of autonomous driving
From the perspective of mass production rhythm, concentration is not dominant. According to the official plan, the production version of Jidu ROBO-01 will be officially unveiled during this year’s Guangzhou Auto Show, and the delivery time is in the second half of 2023; in contrast, Wei Xiaoli has already launched products based on the second-generation platform, with annual sales It will also exceed 100,000; data shows that in September this year, China’s new energy vehicle sales reached 708,000, a year-on-year increase of 93.9%, and the market penetration rate reached 27.1%.
As a “latecomer” in the smart electric vehicle industry, how can Jidu create a differentiated experience when the industry is getting more and more convoluted, and then have the ability to deliver 800,000 vehicle products throughout the year in 2028?
The set degree gives a slightly unfamiliar concept, “automobile robot”. To put it simply, since its inception, Jidu has not only provided simple means of transportation, but also a “car robot” that connects the digital world and the real world. So how do you specifically understand “automobile robots”? Li Yanhong and Xia Yiping have mentioned three words on multiple occasions, namely “free movement”, “natural communication” and “self-growth”.
These words may be a bit abstract. If you break it down into detail, the car robot has high-level autonomous driving capabilities and can move freely; the precise recognition of human-vehicle interaction and speech semantics allows vehicles and people to “naturally communicate”, understand user emotions, and real-time Responding to user needs; automotive robots can learn and iterate themselves according to the user’s usage habits, continuously optimize the experience of autonomous driving and intelligent cockpit, and achieve “self-growth”.
Among them, it can provide continuous high-level intelligent driving from point to point, which is considered to be one of the core competitiveness of Jidu. According to the concentration planning, the point-to-point pilot assist PPA, with high-level intelligent driving capabilities in the “three-domain integration” of high-speed domain, urban domain, and parking domain. At present, Jidu has launched the generalization of high-end intelligent driving scenarios in Beijing and Shanghai to ensure that the official version of the PPA gradually covers core cities across the country.
Set up this Flag, how to achieve it? It is understood that the ROBO-01 Moon Probe Limited Edition is equipped with a fairly hard-core configuration, including 2 NVIDIA Orin chips, with a computing power of 508TOPS; it is also equipped with 31 sensors, including 2 128-line high-resolution lidar , 5 millimeter wave radars, 12 ultrasonic radars, 12 cameras.
Although from the industry point of view, the computing power of 508TOPS is not the strongest, but it is also at the top level of the industry. On the other hand, the number of computing power is not directly equivalent to the actual processing power, but also requires stronger algorithms and more data. At the algorithm level, the most concentrated reliance is Baidu Apollo’s high-level autonomous driving capability, which has been widely verified on its Robotaxi, with a total test mileage of more than 36 million kilometers.
However, the set degree is not a direct copy of Apollo’s high-level autonomous driving capabilities, but also requires re-integration and development verification. This is mainly because Apollo and Jidu have different positioning and problem solving: Apollo’s goal is to do Robotaxi, which needs to verify the ability of unmanned autonomous driving and solve commercialization problems; in contrast, Jidu is for mass production and C end, aiming at high-end autonomous driving product experience. It is understood that there are currently about 300 or 400 people in Jidu’s autonomous driving department, plus about 1,000 people in Baidu, who are carrying out large-scale integration and verification of autonomous driving.
At the press conference, Jidu showed a video of the autonomous driving test of ROBO-01. In the video, the vehicle can realize unprotected left turn, avoidance in the lane, avoidance of pedestrians, high-speed road sections, recognition of oncoming vehicles, high-speed cross-lane avoidance and other actions. According to Xia Yiping, the current high-end intelligent driving road test version is only equivalent to about 50% of the mass-produced version, and the autonomous driving capability will be further improved in the future.
Voice interaction experience
In Jidu’s view, in order to truly become an “automobile robot”, in addition to free movement through high-level autonomous driving, it is very important to achieve precise recognition of human-vehicle interaction and speech semantics.
Before that, since speech recognition requires a lot of computing power, most of the instructions need to be uploaded to the cloud server for calculation and implementation, which will not only cause a certain delay (usually 1.5 seconds), but also will In mountainous areas and other environments with no network and weak network, the voice function is unavailable or restricted.
In order to improve the user’s interactive experience, the industry generally believes that multimodal interaction is a feasible and efficient option. The so-called “multimodal interaction” is a combination of human senses such as vision, hearing, touch, etc., and the computer uses a variety of communication channels to respond to input and fully simulate the interaction between people.
However, Jidu has its own unique understanding of scenarios and needs, and believes that voice interaction is the most important interaction method for smart cockpits, and the experience can be greatly improved. Xia Yiping once said, “We are not doing simple instructions, but really achieve very natural language communication. If you want to recognize a simple statement, such as closing or opening a window, you can do a script locally, but you need to recognize a Natural language, if you go to the cloud, you have to go around the circle when you come back, and it is also affected by the network connection, such as trying it in the basement.”
To this end, Jidu ROBO-01 Moon Probe Limited Edition uses Qualcomm 8295 chip, with a computing power of 30TOPs, which is 8 times that of the current flagship chip Qualcomm 8155. This is also the first model in China with the 8295 chip.
Relying on the powerful computing power of 8295, Jidu ROBO-01 can put the speech recognition algorithm model in the cloud directly on the car side, which can increase the real-time response speed from 1.5 seconds to 700 milliseconds, and improve the single-person continuous voice popular in the industry. Extended to multi-person continuous voice. Full-featured speech recognition capabilities can be achieved even when offline, which greatly expands the usage scenarios of speech recognition and prevents artificial intelligence from becoming “artificial mental retardation”.
At the same time, by carrying Qualcomm 8295 and the intelligent architecture JET, Jidu has realized the fusion backup of cabin and driving. Simply put, both the smart driving chip and the smart cabin chip can be used as the basic computing power for autonomous driving. If the dual ORIN chips of the intelligent driving domain controller are stuck or crashed, the 8295 chip of the intelligent cabin domain controller can also be installed immediately, which is equivalent to an extra “protective umbrella”.
In addition to the key voice interaction, Jidu ROBO-01 also has the “human-like” interaction capabilities of multi-modal integration such as visual perception, voiceprint recognition, and lip language capture, and also allows people to communicate with the car “naturally and smoothly”.
In addition to “listening”, the limited edition of Jidu ROBO-01 also has a strong “expression ability”. In terms of images, ROBO-01 is equipped with a 35.6-inch 6K ultra-clear all-in-one screen. From the perspective of screen size alone, the concentration is not the largest, but the solution adopted by most car companies consists of several multi-screens. In contrast, Jidu uses an integrated screen, which is three times the screen width of the Tesla Model Y. It can not only meet the user’s driving navigation, game entertainment, online office and other scene-based needs, but also support switching “3D smart phones”. The “Driving Map” mode allows the perception of the car robot to run on the 3D map, creating an immersive 3D panoramic experience.
A very important point is that the intelligent system of Jidu is not “futures”, but “spots”. As the trend of “software-defined cars” becomes more and more clear, “hardware pre-embedded + software OTA upgrade” has gradually become the mainstream strategy of current car companies. However, Jidu solemnly announced that when the first model is delivered in 2023, users will be able to experience the full version of autonomous driving functions during test drives, instead of going through OTA after the launch.
As the first model of a new brand, the strategic significance of ROBO-01 to Jidu is self-evident. This time, the limited edition of ROBO-01 moon exploration is launched, and more of it is to establish a brand, which can be seen from the price of nearly 400,000 and only 2,000 places.
At present, unlike Wei Xiaoli who started from 0, on the basis of Baidu Apollo and Geely SEA, Jidu not only has a good start, but is also on the right path. From the ROBO-01 Moon Exploration Limited Edition, we can see Jidu’s thinking about the shape of a real smart electric vehicle. This thinking is not only about the appearance, but more about the structure and experience.
Next, the focus of Jidu will be the mass production of the first car. Xia Yiping hopes to achieve the “small goal” of monthly sales of over 10,000. After all, no matter how beautiful the “mission” or how advanced the technology is, it ultimately needs to be supported and verified by sales.