LiDAR is accelerating its expansion, and the number of NOA vehicles in China will exceed 1.8 million in 2024

According to statistics from Gasgoo Automotive Research Institute, in the field of driving ADAS, the market share of local suppliers will increase to 14.4% in 2023, and the market share of local suppliers in parking will exceed 30%.

Integrated driving and parking is one of the important solutions in the current intelligent driving field. Today, the market share of integrated driving and parking developed and manufactured by OEMs is close to 50%. Among them, local suppliers are leading the foreign tier with their full-stack software and hardware development capabilities, rapid product iteration, and multi-chip platform adaptation.

From the perspective of components, in addition to lidar, some domestic suppliers have stood out in areas such as air suspension and high-precision maps, while some domestic companies have begun to achieve large-scale supporting development in areas such as intelligent driving domain control chips and wire control braking.

Relatively speaking, there is still a lot of room for growth in areas such as domestic cameras and millimeter-wave radars.

In the field of high-end intelligent driving, the domestic supply chain has developed rapidly, with many outstanding domestic manufacturers emerging from software algorithms, maps, chips to overall solutions.

It has to be said that the mass production of Urban NOA has, to a large extent, driven the explosion of demand in upstream industrial chains such as high-computing chips, lidar, and cloud services, and a number of suppliers have embarked on the path of listing.

On December 20 last year, Zhixing Technology, a major domestic autonomous driving domain controller provider, was successfully listed on the main board of the Hong Kong Stock Exchange. Earlier this year, RoboSense rang the bell on the Hong Kong Stock Exchange, becoming the world's largest laser radar company by market value. Many other intelligent driving companies have IPO plans. On March 28, Zongmu Technology submitted a prospectus to the Hong Kong Stock Exchange, intending to be listed on the main board.

As for car companies, based on the current trend of intelligent driving, independent car companies are accelerating the research and development and application of intelligent driving technology. In terms of specific routes, on the one hand, powerful OEMs choose independent research and development to maintain technological leadership and independence; on the other hand, most car companies choose to cooperate with suppliers and directly purchase mature solutions to reduce costs and achieve rapid implementation. At present, suppliers who can provide intelligent driving solutions in the market include auto parts giants, technology giants and professional start-ups, which constitute a diversified force in the development of intelligent driving.

With the continuous investment of domestic automakers in technology research and development and the expansion of domestic suppliers' market share, the two promote each other and provide strong support for the rapid development of intelligent driving technology. In the future, this trend is expected to continue to strengthen, further promoting the popularization and application of intelligent driving technology in China and even the global market.

Cockpit integration is gradually mass-produced, and autonomous driving is moving towards end-to-end

When talking about the future of intelligent driving, we cannot avoid the "keywords" such as cockpit integration, high-computing chips, big models, and end-to-end.

The concepts of driving and parking integration and cabin-parking integration are gradually maturing, and will eventually move towards cabin-driving integration or cabin-driving integration. The electronic/electrical architecture of automakers is upgrading to a domain-centralized architecture, and the shared technology of driving and parking sensors and domain controllers is becoming more mature.

The development of cockpit fusion technology is in a positive transformation period. Currently, automobile companies and suppliers are planning cockpit fusion domain control, and some companies have launched related products, such as Bosch's central computing platform and Changxing Intelligent Driving's cockpit fusion platform.

It is expected that cockpit fusion domain control will achieve small-scale mass production by 2024. With the maturity of technology and market demand, it is expected to be widely used in the near future.

High-computing-power chips are crucial to achieving cockpit fusion, supporting increasingly complex algorithms and processing large amounts of data, as well as ensuring the safety and reliability of the system.

However, based on the fact that the integrated driving and parking technology is one of the main focuses of current intelligent driving, most of the low- and medium-computing power integrated driving and parking solutions currently installed on vehicles adopt multi-SOC solutions, and the single-SOC solution is adopted on the new-generation platforms. The high-computing power integrated driving and parking domain control mainly adopts multi-SOC solutions because it needs to support higher-level intelligent driving functions and takes into account the immaturity of pre-embedded computing power and system redundancy.

With the improvement of AI algorithms and chip design capabilities, even medium- and low-computing SOC platforms can provide sufficient performance to support the basic functions of intelligent driving. Cost-effective single SOC medium- and low-computing platforms are expected to be the first to explode.

Another crucial influencing factor is the iterative innovation of autonomous driving algorithm technology. Many leading car companies have chosen to follow Tesla's algorithm iteration method, which is mainly divided into three stages:

The first stage: Tesla introduced BEV and Transformer technologies in its autonomous driving system to achieve "free from high-precision maps", which also marked the beginning of the application of large models in vehicle systems; the second stage: upgrading the occupied network to achieve "free from lidar"; the third stage is end-to-end autonomous driving.

On March 18 this year, Tesla began to fully push the FSD (full self-driving) V12.3 version in North America. It is worth noting that the V12 version introduced the "end-to-end neural network" technology, which once again sparked heated discussions in the industry.

The current intelligent driving system adopts a modular model, which divides perception, prediction and planning into three independent modules, and the technology stack of each module is quite different. This modular architecture may lead to information loss during information transmission, making the system complex and difficult to maintain, and unable to effectively cope with complex road environments.

In contrast, the end-to-end model has obvious advantages in intelligent driving solutions. It integrates perception, prediction and planning into a single model, simplifies the structure of the solution and improves computational efficiency. Compared with traditional rule-based models, the end-to-end model is easier to scale and achieve performance breakthroughs.

However, the disadvantages of the current end-to-end model are also obvious. End-to-end models are often viewed as "black boxes" that combine multiple steps into one model, making it more difficult to understand how the model works internally. This opacity may lead to problems with interpretability and explainability. Moreover, end-to-end models require more computing resources, greatly increasing the cost and time of training and deploying models. In addition, end-to-end models may also face problems such as difficulty in debugging and optimization, and over-generalization.

However, due to its obvious advantages and outstanding features, it does not prevent many companies from believing that end-to-end autonomous driving is one of the most promising ways to achieve driverless driving in the future. Many companies such as NIO, Xiaoli, and Yuanrong Qixing have chosen to follow Tesla's technology route and deepen their exploration on this path.

The autonomous driving industry has reached an inflection point and we are likely to see more exciting innovations and advances in the coming years.