Shulou(Shulou.com)11/24 Report--

In today's automobile industry, the competition for intelligent technology has become white-hot. According to the survey, 84.5% of consumers said that the intelligent function significantly affected their car purchase decisions. Especially intelligent driving and intelligent cockpit, these two technologies have become the two most eye-catching pearls in the smart car industry and the inevitable competition in the market.

At the Baidu World Congress on October 17, Ji Yue publicly demonstrated the intelligence module interaction ability of all intelligent cars in the world, and released the first domestic BEV+Transformer "pure vision" high-level intelligent driving technology. At the same time, the first domestic OCC (Occupancy Network) jointly developed by Ji Yue and Baidu has also been announced for the first time.

From the perspective of the new technology display, Ji Yue's technology is not only at the forefront of the industry again, but also will lead the industry trend.

The pioneer of the high-level intelligent driving scheme, the first domestic BEV+Transformer high-level intelligent driving was released.

According to reports, the first domestic BEV+Transformer "pure vision" high-level intelligent driving technology program carried by Ji Yue 01 has achieved pilot-assisted driving on urban roads in Shanghai. To conquer the busy city of Shanghai with such complex road conditions, it is believed that the full landing of this technology has entered the timetable, and the popularity of urban smart driving in China as a whole will be accelerated.

In a video of a city pilot-assisted driving test posted on the official social account of Ji Yue, a Ji Yue 01 with only a pure visual perception structure ran through Shanghai's Lujiazui-Bund-Nanpu Bridge and other sections of the road, achieving zero takeover under complex road conditions.

This means that the point-to-point navigation assistance PPA, the core function of Ji Yue's advanced intelligent driving assistance system, has run through the urban roads of Shanghai with pure visual ability, becoming the world's first "pure visual" high-level intelligent driving scheme in urban areas.

At present, Jiyue is the first and only smart car brand in China to adopt the "pure vision" smart driving scheme. There are a total of more than a dozen cameras in the perceptual architecture of Ji Yue. Among them, there are seven 8-megapixel high-definition cameras, which is also the largest number of smart cars.

With the release of this "pure vision" advanced intelligent driving technology plan, Ji Yue also came up with another important "cool techs"-the first domestic OCC (Occupancy Network) jointly developed with Baidu to occupy the Internet technology. This is the first time that the technology has been announced to the public, which not only fills the gaps in related technologies, but also contributes a new solution to the world problem of high-level intelligent driving technology.

In other words, Ji Yue's high-level smart driving will eventually form a complete technical framework based on BEV+Transformer "pure vision" and OCC occupying network technology.

Under this BEV+Transformer "pure vision" + OCC technology scheme, the information collected by high-definition cameras is more abundant than the sparse and discontinuous point clouds generated by lidar, which can integrate 3D geometric information with semantic information and help automobile robots restore 3D scenes more accurately.

In addition, with the reduction of hardware of "pure vision" intelligent driving technology, the model is more efficient and centralized, and it also has more advantages in technical iteration, data closed loop and so on. The quality and efficiency of data is the key to the decisive victory of intelligent cars.

Under the BEV+Transformer "pure vision" scheme, cars can quickly and accurately predict the spatio-temporal position and attitude trajectories of road participants, and provide sufficient and accurate data support for further decision-making.

In the traditional perceptual recognition technology, because the actual driving situation is very complex, there are a variety of corner case. It is difficult for enterprises to be exhausted one by one in the scene, which brings a difficult problem to the landing of the smart driver. On the other hand, this set of technology changes the way of thinking, makes the perception of the car simple and accurate, fundamentally avoids the problem of missed detection of objects in the non-training set by traditional vision, and greatly improves the "pure vision" high-order intelligent driving ability of the automobile robot. In the future, this technology may trigger a new wave of domestic smart driving technology.

Specifically, the technology is based on visual signals and understands and processes spatial information in a voxel way. To put it simply, with the support of this set of technical tests, when a car finds an object in the surrounding environment, it only needs to consider whether the voxel is occupied, but not what the object is. This characteristic makes the technology better adapt to different scenes and environments.

The countdown to the "boarding" of the big model leads the intelligent cockpit to the next level.

With the development of automobile intelligence, the attribute of car as a pure means of transportation is being deeply subverted. Intelligent car is occupying the living space of more and more people, and some people even call it the "third space" outside work and home.

Affected by this, the industry's demand for the intelligence of the automobile is also improved, in which the ability of the intelligent cockpit is especially valued by users. In the cockpit, whether the car can accurately and quickly identify the complex voice instructions of drivers and passengers has become one of the core competencies of automobile manufacturers, and it is also the standard to judge whether they are ahead of the industry.

In the iterative process of intelligent cockpit, the gradual reduction of physical keys is a major technical trend of the current development. For example, the traditional dashboard is replaced by HUD, and a number of cumbersome buttons are "brought" into a simple central control screen. Ji Yue 01 uses a 35.6-inch 6K ultra-clear all-in-one screen and removes physical buttons as much as possible.

This time, Ji pushed this trend to a new height and achieved the extreme in the industry. Ji Yue demonstrated to the outside world the world's leading intelligence module interaction ability, all the functions of the vehicle can be controlled by voice, and there are almost no physical buttons in the cockpit. It is worth mentioning that it is the first to achieve the full localization of the voice AI algorithm, the fastest global full offline speech recognition speed reaches 500ms, and supports functions such as no wake-up for the main driver and simultaneous chat in the four-tone area, which can still achieve the highest speed of interaction without the network.

At the 2023 Baidu World Congress, Li Yanhong showed to the outside world that the voice interaction of Ji Yue 01 can not only cover complex scenes such as inside and outside the car, four-way theory, continuous dialogue, etc., but also achieve the integration of voice, gestures, visual recognition and other multimodal interactions to achieve "fast, accurate and stable". With only one SIMO, vague instructions can also get smart interaction results.

It is reported that Ji Yue 01 will be launched with Qualcomm 4th generation Snapdragon digital cockpit platform 8295 chip, and for the first time fully enable dual NPU (60TOPS). With the support of powerful computing power, it can achieve the smooth interaction of the 35.6-inch 6K ultra-clear screen and put an end to the situation of Catton.

Voice interaction is considered to be the base of the large model landing module, and the excellent hardware also provides the AI computing basis for the large model to "get on the bus". At present, the exploration and application of AI model has been carried out in Jiyue. Under this advantage, the iterative speed of Jiyue has been further improved and the leading position in the industry has been opened. As the first product of Ji Yue, Ji Yue 01 will also be equipped with a large car-side voice model in the future to achieve further progress towards "natural communication".

At present, the pre-sale of Ji Yue 01 started on September 19, with a pre-price of 259900 yuan. According to the official introduction of Ji Yue 01, Ji Yue 01 will be officially launched at the end of October and start delivery at the same time, with all high-level intelligent capabilities "out of the box".

At present, intelligent capability has been an important selling point for almost all automakers, but some of them are often not fully developed, and there is a long way to go. However, all the high-level intelligence capabilities of Ji Yue are "out of the box". This is commendable in the current automobile consumer market.

On October 8, the first batch of Ji Yue 01 mass production cars have been shipped to the National offline experience Center, and the full-function test driving experience will be launched soon. As the pioneer of the domestic "pure visual" smart driving program, at the beginning of its debut, he chose the busy city of Shanghai with complex road conditions as the "assessment" standard, which is enough to see the strength and confidence of Ji Yue. In the highly competitive Chinese car market, what kind of storm will be caused by Ji Yue 01 is worth looking forward to.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.