On the same day, Xiaopeng, GAC, and Jike all announced the landing of L3 this year. The 2025 Wanjie M9 and Xiangjie Extended Range Edition released by Hongmeng Zhixing on March 20th will also be star models challenging the landing of L3. And Jike, who started relatively late by one position, is catching up. In the recent parking spot to parking spot preliminary race, I saw Jike's figure this time.

Jike previously announced that it will launch the "parking space to parking space" intelligent driving function in the first quarter of 2025, and the core purpose of the Jike Qianli Haohan Intelligent Driving Conference on March 18th is to "fulfill this promise".

This commitment includes not only the parking space to parking space in the present time, but also the L3 in the future time. Jike has also joined the most intense and challenging L3 landing battle in the intelligent driving war in a short period of time. one

What is a fully blooded version of parking space to parking space

In less than a month, Jike will enter the next stage: in April, Jike's "parking space to parking space" intelligent driving function will be launched on a large scale, equipped with Jike Qianli Haohan Intelligent Driving H7 and above models.

In early March, Geely Holding Group released its self-developed intelligent driving solution "Qianli Haohan Intelligent Driving", which unified the intelligent driving solutions of Geely's various brands and installed them in different models according to the H1, H3, H5, H7, and H9 hierarchical levels (see "Geely launches the intelligent driving defense war" for details).

Among the five versions of intelligent driving solutions H1, H3, H5, H7, and H9, H7 and H9 are the highest level two solutions. At the same time, Jike, as a high-end new energy brand under Geely, also shoulders the responsibility of launching advanced technology.

Therefore, in the description of Jike, the "parking space to parking space" in the Jike Qianli Haohan Intelligent Driving H7 solution has an additional prefix of three words compared to its competitors - "full health version".

The full blood version can be understood as not only a marketing statement of technical integrity, but also reflects its comprehensive investment in hardware embedding, data training, and scene coverage, aiming to make car owners feel a "castration free experience". Chen Qi, Chief Scientist of Intelligent Driving at Geely Holding Group and Vice President of Jike Group, introduced that the so-called "full blood version" refers to the coverage of all scenarios and nodes from parking space to parking space, and there is no need to learn about parking lots. "You can go to the whole country, park as soon as you have a seat, and use it for the first time without taking another step. Jike's "first time use" corresponds to the fact that some car manufacturers' "parking to parking" solutions currently require memorizing routes, because the general parking to parking function requires car owners to train it once in advance.

But it's not that Extreme Krypton's "parking to parking" system completely eliminates the need for memorization, but rather that in certain situations, it can be used without memorizing the route for the first time.

The premise is that the destination set by the car owner is the supermarket or park parking lot where Jike's Jike charging station is located. These areas have already completed the mapping work ahead of schedule by Jike, so the intelligent driving system can go and park in the parking space on its own.

In the current reality, many companies' city navigation and even parking to parking intelligent driving still have some scenarios that are not covered, such as roundabouts or U-turns that require exiting. However, this version of Jike is based on H7's "parking to parking" feature, seamlessly connecting the national city map free NZP and parking to parking functions, and providing comprehensive access to key scenarios such as roundabouts, U-turns, variable lanes, highway ETC, and waiting areas at intersections without exiting.

In addition, the press conference also showed that in terms of automatic parking, Jike Intelligent Driving not only covers extremely narrow parking spaces and dead end road parking spaces, but also achieves automatic parking in mechanical parking spaces. When there are no available parking spaces, the 'Find Nearby Parking Space' function can locate the nearest available parking space.

Of course, the prerequisite for achieving the above parking to parking capability is to have redundant hardware pre embedded and sufficient data training, all of which cannot be separated from the efforts made by the Jike self-developed intelligent driving team.

Intelligent driving cannot do without five elements - hardware, algorithms, computing power, and data, but another one that is often overlooked by many technicians is the team.

two

First detailed explanation of MLM, digital foresight network and other technologies behind it

Through our own speed, we not only need to enter the first tier of intelligent driving, but also become a leading intelligent driving enterprise. "This is the core viewpoint of An Conghui in the media interview after the meeting.

An Conghui frankly stated, "The intelligent driving team under the leadership of Da Bai (Chen Qi) has seen our speed, and I believe that in the future, our iteration speed will also be faster

Iteration is reflected in the algorithm, which is an upgrade of the end-to-end Plus technology architecture of Jike Intelligent Driving.

Specifically, end-to-end Plus has evolved from a two-stage approach to a one-stage end-to-end approach, packaging the entire process of perception, prediction, decision-making, and planning into a whole, maximizing the simulation and restoration of the driving style of excellent human drivers. Moreover, it relies on both the "MLM Multimodal Big Prophecy Model" and the "Digital Foresight Network".

Jike's digital foresight network achieves "experiential" decision-making for intelligent driving through the accumulation of spatiotemporal memory and multimodal fusion. Its core concept is similar to the technology of enterprises such as Momenta and Zhiyuan, but Jike emphasizes the systematic integration of predictive capabilities.

In terms of the relationship with VLA and VLM, the concept of digital prerequisite networks does not replace these models, but rather enhances the prediction dimension to form a more complete "perception decision execution" loop. This technological path may provide new possibilities for the implementation of advanced intelligent driving.

In Chen Qi's view, MLM and VLA are not fully conceptual. VLA is equivalent to integrating control and execution, but MLM is multimodal.

Chen Qi summarized the development process of intelligent driving technology as a step-by-step evolution around "experience continuity", which essentially achieves a qualitative change from discrete functions to seamless integration across the entire domain through technological iteration

1. Early stage (function discretization): Independent functions such as AEB (automatic emergency braking), ACC (adaptive cruise control), LCC (lane centering control), etc. are the main focus, with each module triggering logic isolation and frequent operation interruptions (such as the need to reactivate ACC after AEB intervention). 2. Breakthrough in local continuity: High speed NOA (navigation assisted driving) achieves seamless operations of lane keeping, automatic lane changing, and on/off ramps, significantly improving the continuity experience in high-speed scenarios. 3. Global continuity upgrade: The urban NOA is further expanded to complex urban roads, and the "parking to parking" function (D2D, Dock to Dock) completely eliminates scene breakpoints by integrating traffic and parking throughout the entire chain (basement city highway target parking space), marking a systematic solution to experience continuity issues.

Of course, Chen Qi also said that many of the experiences on the platform are well done. "I think everyone is in a process of letting a hundred flowers bloom and a hundred schools of thought contend," he said

Currently, the competition for intelligent driving based on large models is focused on computing power, algorithms, and data. Car companies believe that more and higher quality intelligent driving data can help train large models.

In terms of computing power, Geely recently established the Star Wisdom Computing Center 2.0 with technology ecological partners such as Step Leap Star, Qianli Technology, and Xingji Meizu, with a comprehensive callable computing power of 23.5ELOPS.

On the data front, there is an insurmountable issue of compliance. Both An Conghui and Chen Qi emphasized that 'Geely has surveying qualifications and its own mapping merchants', while currently only Geely and Huawei are domestic automotive related companies with surveying qualifications.

After all, 'having surveying qualifications' is the foundation for building the' digital foresight network 'capability.

In terms of hardware, the Jike Qianli Haohan intelligent driving products on sale now come standard with dual NVIDIA DRIVE AGX Orin-X chips, and the latest flagship model, the Lynk&Co 900, which will be available for pre-sale in March and launched in the second quarter, will be equipped with NVIDIA DRIVE AGX Thor chip and H7 solution for the world's first launch. In an interview, An Conghui and Chen Qi both stated that the Thor chip is currently progressing smoothly in the car, and "as far as I know, there is no shortage problem yet

Last year, Lynk&Co achieved a total sales volume of 285000 vehicles, and its goal for 2025 is to achieve 390000 vehicles. When Lynk&Co merges with Jike and adopts the same intelligent driving system and hardware configuration, it means that while improving economies of scale, it will also provide more data for its self-developed vast intelligent driving, accelerating the iteration of intelligent driving.

The first Lynk&Co model to integrate with the Jike Haohan Intelligent Driving System is the Lynk&Co 900. Jike also revealed that the Lynk&Co 900 will start pre-sales on the 25th of this month and is expected to be launched after the Shanghai Auto Show; The delivery time is tentatively scheduled for the second quarter of this year.

If the H7 intelligent driving solution is the last shot fired by Extreme Krypton in the L2.9999+finale, then the H9 solution is their first shot in welcoming the L3 battle.

three

Shouzheng Yiqi: Approaching the finish line of the L3 battle, the Extreme Krypton 9X Brilliance will be the first model of the H9 scheme, scheduled to make its debut at the Shanghai Auto Show.

Regarding the H9 intelligent driving solution, Jike will introduce more from some underlying solutions and technologies, emphasizing stronger capabilities and higher redundancy. Hardware chips and perception are the biggest highlights.

In terms of hardware infrastructure, Jike has prepared the "Jike Qianli Haohan self-developed intelligent driving domain controller" for L3. It will be equipped with two Thor chips for the first time, with a computing power of 1400 TOPS, which is 175% higher than the industry's top dual NVIDIA DRIVE AGX Orin-X chips.

In terms of perception hardware, the configuration of the Extreme Krypton L3 will also be equipped with 5 laser radars, with 1 as the main radar and the remaining 4 as supplementary blind radars; It is claimed that in car perception, it is necessary to achieve "4 360 °" coverage, in addition to laser radar, millimeter wave radar, visual perception, and ultrasonic radar all need to achieve 360 ° coverage without blind spots.When it comes to LiDAR, Chen Qi said that in some low light or other extreme scenarios, LiDAR can be seen for a long time, and distances of several hundred meters or tens of meters can be life-threatening. Using five LiDAR can mainly provide safety redundancy.

2025 will be the year of the L3 battle for major car companies.

2025 is not only a technological turning point, but also the first year of L3 commercialization. The CICC report points out that policies, software and hardware, and guarantee mechanisms are in place, and car companies need to face the problem of "responsibility attribution": who will pay for system failures? User experience becomes the ultimate battlefield.

Ideal, Xiaopeng, Huawei and other car companies or suppliers are fiercely competing for the commercialization of L3 level intelligent driving. This competition not only concerns the leading technological path, but also directly determines the market share and user mental dominance of enterprises during the regulatory window period.

The goal set by Extreme Krypton is to have the ability to deliver L3 mass-produced vehicles by the end of this year, which is not conservative, but even aggressive. Is it possible for Jike Technology Group to become one of the earliest car companies in the industry to achieve L3 level intelligent driving capabilities?

The Shanghai Auto Show in April this year will be the first time that the Extreme Krypton Super Electric Hybrid 9X and L3 intelligent driving systems have undergone market testing.