On January 14, at the Extreme Day of Evolution of Extreme Automobile Robotics, Xia Yiping, CEO of Jidu and CEO of Extreme, drove the Extreme 01 from Shanghai Jiading to the Extreme Hangzhou City West Intime Experience Store in live broadcast form, using point-to-point navigation assistance driving PPA throughout the journey, and announced that OTA will undergo a major version upgrade. In addition, Xia Yiping also revealed the news about the follow-up intelligent driving plan and the second model Extreme 07.
Ji Yue 01 welcomes its first software upgrade, with over 400 functional points upgraded and iterated. Ji Yue announced that it will undergo its first major version upgrade, gradually pushing V1.3 software to all users via OTA. This is the first vehicle-level upgrade for Ji Yue 01 since its delivery at the end of October 2023. OTA 1.3 will start pushing in batches from the early morning of January 15, and OTA 1.3.1 will start pushing in batches around the time of the Spring Festival. According to the introduction, V1.3 includes the evolution of general visual capabilities including OCC occupying network technology on the car, the evolution of general voice capabilities including the full integration of SIMO into Wenxin Yanyan 4.0, safety and ecological services such as winter mode, and more than 400 functional points upgraded, which will greatly enhance the intelligent experience of Ji Yue 01.
Specifically, the upgraded OCC network technology will be jointly developed by Jiyue and Baidu. After the upgrade, Jiyue will form a complete system of advanced intelligent driving with “BEV+OCC+Transformer”, greatly improving the ability of “pure vision” intelligent driving to recognize irregular obstacles and generalize scenes. Jiyue officially named “BEV+OCC+Transformer” as “B.O.T. Tri-Foil”, representing the upgrade of the world in the eyes of the robot from two dimensions to three dimensions, allowing the car robot to have a sense of “intuition”. The “pure vision” technology solution of BEV+Transformer has been the only mass-produced in China on Jiyue 01.
JiYue stated that Baidu and JiYue have achieved efficient model training and rapid iteration every week based on a high computing power training cluster of nearly 2 EFLOPS. With the empowerment of Baidu’s advanced Apollo autonomous driving technology ecosystem, JiYue relies on the “B.O.T triple foil” perception technology architecture, and the 3D accuracy of the pure visual solution can reach the centimeter level, comparable to or even surpassing LiDAR. The speed recognition accuracy of moving obstacles is also controlled within a range of 0.1 meters per second. In terms of optimizing the point-to-point navigation assistance PPA function, JiYue’s point-to-point navigation assistance PPA can achieve advanced intelligent assistance in multiple scenarios such as high-speed and urban driving. Currently, 90% of the national high-speed and elevated roads can use high-speed PPA, and urban PPA has been tested in the main urban roads of Shanghai, Beijing, Shenzhen, and Hangzhou. “The point-to-point navigation assistance PPA has already covered more than 400,000 kilometers of roads nationwide. During user travel, the PPA smart driving mileage accounts for 2-4 times the total travel mileage of basic assisted smart driving by competitors, with Shanghai even reaching as high as 47.9%,” Xia Yiping added.
Optimize driving details to enhance the driving experience This OTA upgrade will upgrade the Wenxin Yiyandam model from 3.5 to 4.0, and the intelligent AI partner SIMO will also be fully integrated with Wenxin Yiyandam. Users can directly ask SIMO to experience all the capabilities of Wenxin Yiyandam 4.0 through voice, and still support other functional commands while using Wenxin Yiyandam. SIMO supports querying the number of traffic lights within 5 kilometers, offline querying of car temperature/date, asking for the current singer/song name, and can customize turning on/off other SIMO voice command recognition areas. In future version upgrades, Baidu Maps and Xiaodu will also be fully integrated with SIMO. For example, SIMO will be fully integrated with Xiaodu, allowing control of Xiaodu’s smart home devices and access to over 1500+ ecological brands, truly achieving mutual control between the car and home. In terms of cabin entertainment ecosystem services, new features such as Bilibili, Xiaoyuzhou, and rear-seat karaoke have been added, as well as support for mobile and Switch screen casting, microphone, and virtual joystick control for in-car gaming through the Ji Yue app.
In addition, this OTA update focuses on the details of high-frequency user scenarios. For example, adding remote steering wheel heating, remote defogging and defrosting, automatic window closing on rainy days, remote reservation of preset cabin environment, battery preheating when navigating to a fast charging station, pre-departure battery heating; adding practical functions such as automatic turn signal extinguishing, automatic window closing on rainy days, seamless switch of phone audio, and blind spot video display. At the same time, iterative optimizations have been made to address previous user complaints about navigation positioning, parking space voice optimization, driving recorder optimization, and HMI optimization. Xia Yiping emphasized that the biggest experience of this upgrade is the guarantee of safety, “safety must always come first,” which is also the key focus of the Jiyue car robot development. In order to ensure safety in extreme scenarios, this OTA update also includes a series of deep optimizations. For example, in the event of a collision and power failure, rescuers can press the door open button outside the vehicle to activate the “ice-breaking function” by applying additional force to open the door quickly in extreme situations such as slight deformation of the door or insufficient sealing force, ensuring successful rescue. In addition, the new snow assistance function can optimize the performance of the vehicle in snow scenes, improve its ability to escape, and effectively enhance driving safety. Jiyue’s intelligent driving plan will be launched in 200 cities in China this year. There has always been a debate in the market between lidar and pure visual solutions. Xia Yiping believes that the use of vision to replace lidar in the industry is because the latter is too expensive, but as it gradually moves towards pure visual solutions, it will be found that “vision itself is not a cost issue, it is actually a technical issue.” He said that compared to lidar solutions, pure visual solutions are more difficult, relying on a large amount of data and computational training, and the computational cost is very expensive, but the upper limit is also higher, the model iteration speed is faster, and it can effectively reduce the cost of purchasing advanced intelligent driving experiences for users. Compared to lidar solutions that pass on hardware costs to consumers, pure visual solutions not only place higher demands on enterprise technology research and development, but also mean more “invisible costs” behind the high-intensity research and development investment.
It is worth noting that the current market for smart electric cars has become extremely “overlapping”, the competition among new carmakers is becoming increasingly fierce, and the strength of independent brands is constantly emerging. For example, Huawei’s continuous disruption of intelligence, the newcomer Xiaomi car is also expected to officially debut in 2024, which will further intensify the market competition. Xia Yiping expressed that the current three-electric field has been fully developed, and 2024 will definitely be a very developed year for intelligence. As for the opening of intelligent driving in a city, in his view, a city must achieve a coverage rate of sixty to seventy percent or more to be considered open. Jiyue expects to launch intelligent driving in 200 cities in China this year. Regarding the current debate between light maps and no maps, he stated that there is currently no real “no map” because the current computing power cannot handle it, and everyone’s definition of “no map” is different. Overall, “no map” is a gradual iterative process and the ultimate goal. According to the plan, Jiyue is expected to launch light map intelligent driving at the end of the second quarter or the beginning of the third quarter this year. In addition, regarding the release of the next generation of vehicles, Jiyue 07, Xia Yiping stated that it can be expected at the Beijing Auto Show. At that time, the intelligent cockpit and brake driving of the vehicle will be upgraded to version 2.0.