Xia Yiping: 2023 will be the first year of competition in automotive intelligence

From March 25 to 27, 2022, the China Electric Vehicle Hundred People Forum (2022) was officially held. The theme of this forum was "Welcome to the New Stage of Market Development of New Energy Vehicles", focusing on topics such as the new energy vehicle market structure, the coordinated development of smart cities, and the new direction of smart electric vehicle consumption. The conference widely invited leaders of relevant domestic ministries and commissions, representatives of foreign government agencies, industry-leading companies, experts and scholars to give speeches. Autohome has excerpted some of the constructive views and shared them with readers.

During the meeting, Jidu CEO Xia Yiping shared his views on building new strengths in the era of smart cars.

The following is the transcript of the speech (condensed):

Jidu just celebrated its first birthday in early March. In fact, Jidu is a new face to the entire automotive industry and everyone here.

The theme of this year's forum is "Building New Strength in the Era of Smart Cars", which fits Jidu's brand and product concept very well. Jidu is going to build a car robot product. Why do we call it a robot? Because when a car has a high degree of intelligent interaction and high-level autonomous driving capabilities, it is actually a robot in the form of a car - Jidu's products will be driven by AI, and provide users with a high-level autonomous driving product experience while meeting regulations, and also have the ability to continuously iterate.

Today, I am very honored to discuss with colleagues in the industry the mass production and popularization of cutting-edge intelligent technology and usher in the era of smart cars 3.0.

In just over a year since Jidu was founded, I have actually deeply felt that the automotive industry is undergoing an epoch-making innovation. So what are the products of the smart car 3.0 era? In the era of fuel vehicles, when a person stays in the car to rest, he has to start the engine to enjoy the air conditioning in the car and the media services brought by the large screen in the middle for a long time. This is actually very inconvenient, unsafe, and not environmentally friendly. But now, electric cars can also use air conditioning and watch movies when they are stationary, and people are gradually staying in the car longer when they are stationary. In the era of fuel vehicles, the tool attribute of the car is stronger, and it is more of a tool from point A to point B; but now, with the popularization of electric vehicles, smart electric vehicles are becoming our second living space.

On the other hand, electric vehicles are experiencing a comprehensive digital upgrade, with AI becoming a new trend. However, for a long time, the computing power of automotive chips has been lower than that of smartphones, which has resulted in many AI technologies not being able to play their advantages in cars and not being able to fully improve user experience. But when it comes to smart cars 3.0, we have given cars enough computing power. The emergence of high-computing automotive chips such as Qualcomm 8295, NVIDIA, and Horizon Robotics has gradually transformed cars, a means of transportation, into an AI-driven smart mobile space.

AI has brought about technological innovation, efficiency improvement, and subversive experience. We believe that 2023 will be the first year of automotive intelligence competition, and the era of smart cars 3.0 has arrived! So, how should we define the products of the smart car 3.0 era? I think there are the following points:

First, it has software and hardware systems that support advanced autonomous driving capabilities. Whether in cities or on highways, the car's autonomous driving has grown from being able to cope with only specific scenarios to being able to cope with more complex scenarios with ease.

Second, voice and semantic interaction must be accurately recognized and the communication process must be natural.

Third, it can also complete autonomous learning and function iteration based on the user's car usage habits. Whether it is autonomous driving or voice interaction, as the user accumulates time using the car, it can automatically optimize the corresponding functional experience, allowing users to drive better and better when using the car, and the functions in the car will understand the user better and better.

Behind these experiences and capabilities, there is actually a whole set of complex software and algorithms, as well as hardware that meets the corresponding computing power requirements, which can be jointly supported to achieve them. To this end, Jidu has developed its own electronic and electrical architecture and domain controller to fully ensure the security of the overall architecture from the perspective of software and hardware integration. Therefore, software development, software capability verification, and the safety and stability of software operation have become the core of Jidu's attention in the era of smart cars 3.0.

AI requires the integration of hardware and software. Hardware security comes from quality, system and appropriate computing power redundancy. Software security comes from the accumulation of big data in a wider range of scenarios and faster iteration of algorithms.

In the era of smart cars 3.0, both autonomous driving and the AI ​​capabilities of smart cockpits place high demands on in-vehicle computing power. The redundancy of computing power improves the certainty of the implementation of high-level autonomous driving capabilities. At the same time, in order to achieve the integrity and stability of high-level autonomous driving algorithms, high-definition cameras, millimeter-wave radars, and lidars will become standard hardware for high-level autonomous driving smart cars. In addition, domain controllers, smart chassis and other configurations for high-level autonomous driving will also become standard for smart cars. At present, Jidu is also developing its own wire-controlled chassis technology, with the aim of ensuring the high requirements of high-level autonomous driving for the integration of software and hardware.

The safety and stability of the operation of AI smart driving and smart cabin software requires not only the guarantee of high-level hardware, but also efforts in software algorithms and the accumulation of data through safe driving mileage to make car robots smarter.

Jidu is the only smart car brand in the industry that fully applies Baidu Apollo's advanced autonomous driving capabilities. Apollo's autonomous driving capabilities have been widely verified on its Robotaxi, and Jidu's products have inherited Apollo's full set of capabilities and safety systems for advanced autonomous driving. Baidu Apollo's Robotaxi has advanced autonomous driving capabilities and has 25 million kilometers of autonomous driving test mileage with zero accidents. A large number of Robotaxi road test fleets are continuously conducting real road tests in more than 30 cities across the country, and algorithms are continuously iterated based on data accumulation. The purpose is actually to enable this software and the hardware it controls to be integrated and perfectly coordinated with each other, to achieve accurate judgment of vehicle driving conditions, and ultimately to ensure the safety and stability of the operation of advanced autonomous driving vehicles.

Jidu is also the only smart car brand in the industry that integrates and develops the autonomous driving technology verified by Apollo Robotaxi, and brings the high-level autonomous driving experience to ordinary family users. Here I would like to emphasize that Jidu's application of Apollo's high-level autonomous driving capabilities is actually different from Robotaxi, and the problems they solve are also different: Robotaxi solves the problem of early verification and commercialization of unmanned autonomous driving capabilities; while Jidu is facing mass production and consumers, with the goal of high-level autonomous driving product experience, and reintegrates and verifies the high-level autonomous driving capabilities of Apollo. Thanks to the accumulation of tens of millions of kilometers of road test data of Robotaxi, the software development process of Jidu's automotive robots can always remain efficient. To this end, Jidu has also developed a set of automotive robot "brain" R&D processes, which we call SIMUcar software integration prototypes, to decouple the software and hardware development of the entire vehicle, without interfering with each other, and with two tracks in parallel. In this way, the development of high-level autonomous driving capabilities can be advanced, thereby exchanging more sufficient time for the "safety and stability verification" of the software. At the end of last year, the development of our high-level autonomous driving capabilities had achieved dual-domain integration of highways and urban areas.

Here, I would also like to share with you the research and development achievements of our SIMUCar:

Our SIMUCar started running in September last year; in November, dynamic testing was launched, and the development of intelligent driving capabilities was simultaneously realized, and it was updated once a week like the development of smartphones. The reliability of functions such as signal light recognition, right turn at intersections, and night driving was successfully verified. In December, SIMUCar updated the common autonomous driving scene applications in urban and highway domains such as left turn at intersections and autonomous lane change. Last month, our SIMUCar was upgraded to version 2.0. At this stage, we completed the integration of high-level autonomous driving software, vehicle electronic and electrical architecture, and SOA. Everything revolves around the development of the safety and stability of the software and hardware systems, and is smoothly entering mass production. In the future, it will be able to meet more and more complex driving scenarios.