Most of the computing capabilities of a single chip will be assigned to the field of autonomous driving, while meeting the needs of cockpit interactions and the needs of internal and external vehicle linkage, and realize the safety and trustwaged of human -machine sharing driving. In order to meet the needs of these autonomous driving and intelligent cockpit interactions, the horizon adheres to the chip design "from software to software" at the beginning of the chip design, we will consider the calculation tasks required for autonomous driving and intelligent interaction applications, and in the chip from design to design to In the two or three years of large -scale production, what changes will occur in the structure of the mainstream algorithm and how to implement the effective calculation of these architectures in the chip.

Although the complete L5 -level autonomous driving cannot be commercialized in the short term, the related technologies of smart cars, including assisting driving and smart cockpit, are gradually mass -produced on the car.

After several years of intelligent development of automobiles, smart cockpies have entered a period of rapid penetration. According to the data from the High-end Smart Automobile Research Institute, from January to June 2022, in the new type of passenger cars in the Chinese market (excluding import and export), smart cars (including network) OTA functions). Thousands of vehicles. 41%, the front installation rate is 48%.71%.

While improving the consumer's intelligent driving experience, autonomous driving and smart cockps are accelerating the relationship between people and cars. In the future, smart cars will also expand the direction of driving in human -computer, which will put forward higher requirements for the underlying computing platform and perception algorithm capabilities. In the field of AI chips in high -performance cars, in addition to Nvidia, Qualcomm and other international players, a number of rising rises represented by Horizon also appeared in the Chinese market.

The horizon believes that vehicle intelligence in the entire scene is the best way to achieve human -machine driving. Compared with other players, the horizon will rely on a deeper understanding of Chinese scenes, forward -looking in -depth integration of the industry's leading software algorithm and chip design capabilities, providing the industry's intelligent technology and solutions for vehicles in the industry, and "journey" from the bottom one at the bottom one The series of intelligent driving chips, reaches AI algorithms, tool chain platforms, large -scale production -level autonomous driving solutions, human -computer interaction solutions, etc., and open the collaborative industry players to accelerate the advent of the smart era.

The comfort and safety of human -machine shared driving depends on the establishment of security trust relationships between man -machine. The establishment of this relationship requires a comprehensive perception ability and a more unified computing ability, which involves the high support of the computing platform for large -scale neural networks. At present, in the field of vehicle, there are still two A3955SB chips for autonomous driving and smart cockpit IVIs, but with the advent of the human -machine era, AI computing power will become more and more concentrated. Most of the computing capabilities of a single chip will be assigned to the field of autonomous driving, while meeting the needs of cockpit interactions and the needs of internal and external vehicle linkage, and realize the safety and trustwaged of human -machine sharing driving. This is the concept of vehicle intelligence in the entire scene.

In order to meet the needs of these autonomous driving and intelligent cockpit interactions, the horizon adheres to the chip design "from software to software" at the beginning of the chip design, we will consider the calculation tasks required for autonomous driving and intelligent interaction applications, and in the chip from design to design to In the two or three years of large -scale production, what changes will occur in the structure of the mainstream algorithm and how to implement the effective calculation of these architectures in the chip. By using software to guide the design thinking of the entire chip, the horizon finally enables the chip to achieve more effective calculations in the smart driving scenario.

The central computing platform consists of the latest journey 5 and the next -generation journey 6 chip, which can meet the full scene computing needs of autonomous driving and smart cockpits, support the perception, prediction and planning of multi -sensor, ensure senior autonomous driving, realize inside and outside the car inside and outside the car Efficient linkage. At the same time, a large -scale computing platform consisting of the journey series chip has a high -security and standardized exclusive domain to meet information security. It can avoid driving safety accidents caused by interference in public areas, avoid possible long -range attackers to issue control instructions to the vehicle, and achieve safer to get safer travel.

On the other hand, in autonomous driving, human -machine interaction is an important part of helping drivers and passengers to establish a sense of trust in autonomous driving. A better human -computer interaction design allows the driver to understand the real -time autonomous state information. If it is recommended that the driver can enter the autonomous driving state, when to exit, adjust the time to adjust the autonomous driving strategy, monitor and alarm drivers The state of fatigue, so as to establish a cooperation boundary and trust between the driver and the autonomous driving system.

This puts forward higher requirements for the functional safety level and performance indicators of the cabin. This fast, low -delay, and high -precision technical solution requires the driver's status and instructions to make the feedback of the driving link more timely and ensure safe control. The first simulation test is difficult to meet the sound or gesture of a single mode. Breaking and prediction can better achieve safety.

The horizon is the first multi -mode voice solution. Through the integrated perception of driver's voice, lips, sight, and gestures, a natural, fast response and high -precision dialogue active interaction is achieved, which greatly reduces the error check rate (especially the car control and driving command is particularly critical). With the support of the Horizon's High Computing Power Computing Platform, this solution can be extended to multi -way and voice interaction. In addition, integrate vehicle perception and scene understanding ability, clever and naturally realize the interactive mapping of the real world and the virtual digital world, bringing unlimited travel services