In recent years, with the popularization of automotive technology, intelligent automotive technology has gradually become the focus of the industry. Intelligence not only represents the inevitable trend of technological development but also brings users a more convenient, safe, and efficient driving experience. The realization of automotive intelligence relies on the deep integration of multiple technological fields, including the energy foundation provided by electrification technology, the rapid iteration of intelligent driving algorithms, the development breakthroughs in computing power chips, and the accumulation of large-scale data. The continuous advancement of these technologies and policy support have driven the rapid penetration of automotive intelligence.

Status of the development of automotive intelligence

1.1 Electrification lays the energy foundation for intelligence

The development of automotive electrification has provided a solid energy foundation for intelligent functions. Electric vehicles use power batteries as the power source, enabling them to provide continuous and stable energy support for a large number of sensors, computing chips, and communication devices. These devices are crucial for realizing intelligent driving and intelligent cockpit functions. According to industry data, in 2023, sales of new energy vehicles in China reached 9.495 million units, an increase of 37.9% year-on-year, with a penetration rate of 31.6%. This trend indicates that as the popularity of electric vehicles continues to increase, intelligent functions are becoming increasingly important in the automotive industry.

The improvement in the cost-effectiveness of new energy vehicles has also greatly promoted the popularization of intelligent technology. In the early stages of electrification, the government incentivized consumers to purchase new energy vehicles through subsidies and tax exemptions. With the maturity of battery technology, the reduction of production costs, and the optimization of vehicle manufacturing processes, the economics of electric vehicles have gradually approached or even surpassed those of traditional fuel vehicles, thereby stimulating market demand. This process not only created conditions for the penetration of intelligent technology but also prompted automakers to introduce intelligent functions to meet consumer needs.

1.2 Dual-wheel drive of intelligent driving and intelligent cockpit

Intelligent driving and intelligent cockpits are the two core directions of the development of automotive intelligence. While enhancing the driving and riding experience, they also bring a new competitive landscape to the industry. Intelligent driving technology reduces driver intervention through automation, improving driving safety and convenience. Meanwhile, the intelligent cockpit enhances the interactivity and comfort of the in-vehicle environment through digitization and informatization.

With the improvement of chip computing power, the maturity of sensor technology, and the rapid iteration of algorithms, more and more automakers are beginning to offer intelligent driving functions as standard equipment. From the initial L2 driving assistance system to today's autonomous driving technology with L3 or even L4 capabilities, intelligent driving is rapidly penetrating into various vehicle models. In terms of intelligent cockpits, with the increasing demand for in-car entertainment and information systems from consumers, functions such as in-car multimedia, voice interaction, and panoramic displays have gradually become standard in the market.

Intelligent driving classification standards

From a commercial perspective, intelligent functions not only enhance the added value of vehicles but also open up new profit models for automakers. Unlike traditional hardware sales models, intelligent functions enable automakers to obtain continuous revenue sources through subsequent software updates and function expansions after vehicle sales through the "hardware embedding + software charging" model. For example, Tesla's FSD (Full Self-Driving) function charges users through a subscription or outright purchase model, with a monthly subscription fee of $99 and an outright purchase cost of up to $8,000. This model not only enhances automakers' profitability but also accelerates the promotion and popularization of intelligent driving technology.

City NOA and vehicle-road coordination: core technical paths for intelligence

2.1 Technical challenges and development status of City NOA

The introduction of NOA (Navigate on Autopilot) technology marks a new stage in the development of intelligent driving. Initially, NOA technology was mainly applied in closed or semi-closed scenarios such as highways, with functions including automatic ramp entry and exit, active lane changes, and overtaking. As technology has gradually matured, NOA technology has expanded from highway scenarios to complex urban road conditions, forming a new technical direction known as City NOA. Compared to highway NOA, City NOA faces significantly greater technical challenges, mainly due to the complexity of roads, the diversity of traffic flow, and the high-precision requirements of the perception system for the environment.

City NOA technology requires vehicles to independently identify traffic lights, pedestrians, non-motorized vehicles, and other traffic elements. It must also possess the ability to handle unprotected left turns, complex intersections, and dynamic obstacles. These requirements place high demands on sensors, computing power, and algorithms. Currently, domestic automakers such as NIO, XPeng, and Li Auto are stepping up their efforts to deploy City NOA technology and view it as one of the core areas of competition in intelligent driving. Data from 2023 shows that the penetration rate of City NOA was 4.8%, which is still relatively low compared to highway NOA. However, with the gradual maturity of technology, the market potential for City NOA is enormous, and it is expected to experience rapid growth in the coming years.

Progress in City NOA by mainstream manufacturers

2.2 Vehicle-road coordination: a powerful supplement to single-vehicle intelligence

Single-vehicle intelligence is the foundation of most current autonomous driving technology routes. It relies on the vehicle's sensors, decision-making systems, and execution systems to achieve environmental perception, path planning, and vehicle control. However, single-vehicle intelligence has some technical bottlenecks, especially in extreme weather and complex road conditions, where relying solely on on-board sensors can be challenging for comprehensive and accurate perception. Here, the introduction of vehicle-road coordination technology provides a powerful supplement to autonomous driving technology.

Vehicle-road coordination achieves more comprehensive perception of road conditions through real-time data interaction between roadside equipment and the in-car system. By installing cameras, radars, and other sensing devices on the roadside, vehicle-road coordination can transmit real-time information on the road to vehicles, helping them better cope with "blind spots" or emergencies. For example, in heavy rain or heavy snow, the performance of on-board sensors can significantly decline, while the vehicle-road coordination system can provide stable environmental information through roadside equipment, enhancing the safety and reliability of the autonomous driving system.

While the concept of vehicle-road coordination can indeed promote the development of intelligent driving, its promotion still faces some challenges, particularly in exploring business models. Currently, the construction of vehicle-road coordination mainly relies on government investment, and the market demand from B-end and C-end users has not been fully unleashed. In the future, how to promote the commercial application of vehicle-road coordination and make it complement single-vehicle intelligence technology will be crucial for industry development.

Technical core of intelligent driving: algorithms, computing power, and data

3.1 Algorithms: from traditional perception to BEV+Transformer

Advancements in algorithms are crucial in the development of intelligent driving. Early autonomous driving perception algorithms primarily relied on the fusion of LiDAR and cameras to perceive the environment through multi-sensor fusion technology. However, this traditional approach was prone to occlusion issues in complex scenarios and required substantial computational resources. In recent years, with breakthroughs in deep learning technology, the new perception algorithm BEV+Transformer has gradually become the industry standard.

BEV+Transformer combines Bird's Eye View (BEV) and the Transformer architecture. Through multi-level feature fusion, it can extract high-precision environmental information from data from multiple sensors. This method not only improves perception accuracy but also solves the limitations of traditional algorithms in handling occlusions. Additionally, BEV+Transformer has the capability for multi-modal fusion, processing data from cameras, radars, and other sensors simultaneously, significantly enhancing the stability and safety of the autonomous driving system.

3.2 Computing power: a core resource for intelligent driving

As autonomous driving technology continues to advance, the demand for computing power in intelligent driving systems is rapidly increasing. Autonomous driving systems need to process massive amounts of sensor data and make real-time decisions through complex algorithms, placing extremely high demands on computing capabilities. Especially in advanced intelligent driving scenarios such as City NOA, the system must not only process dynamic environmental information around the vehicle but also make quick and accurate judgments under complex traffic conditions.

To address this challenge, automakers are deploying high-performance computing chips and establishing supercomputing centers to meet the demand for massive data processing. Currently, NVIDIA's Drive Orin chip has become the preferred choice for high-end vehicle models due to its powerful computing capabilities, providing redundant computing power for future, more advanced autonomous driving functions. Additionally, domestic intelligent driving chip manufacturers are also stepping up their research and development efforts and gradually gaining market recognition. In the future, with the further increase in demand for computing power, deep cooperation between automakers and chip suppliers will become an important driving force for industry development.

3.3 Data: multi-sensor fusion and the evolution of high-definition maps

The operation of autonomous driving systems relies on massive amounts of data obtained from sensors. Currently, intelligent driving systems primarily rely on a combination of cameras, millimeter-wave radars, ultrasonic radars, and LiDARs. Different types of sensors have their unique advantages: cameras can provide 2D visual information, radars can provide information on object distance and speed, and LiDARs can generate 3D environmental models, helping vehicles achieve more precise path planning and autonomous positioning.

High-definition maps are an essential component of autonomous driving, providing precise road environment information, including lane lines and traffic signs. However, the production and maintenance of high-definition maps are costly, and their update cycles are long, posing challenges for large-scale applications. To address this, some automakers have begun exploring solutions that reduce reliance on high-definition maps. By using real-time environmental perception and dynamic mapping technology, they aim to enhance system flexibility and scalability. For example, brands like XPeng and Li Auto have announced plans to gradually reduce their dependence on high-definition maps and instead use real-time perception technology to improve system performance.

Cost optimization and business model exploration

4.1 Cost reduction of LiDAR and the rise of vision-based solutions

LiDAR was once an indispensable sensing device in high-level autonomous driving systems, but its high cost has always been a challenge for automakers. In recent years, with the rise of domestic LiDAR manufacturers and the release of economies of scale, the cost of LiDAR has significantly decreased. For example, domestic manufacturers such as Hesai Technology and RoboSense have become leaders in the global LiDAR market, with LiDAR prices falling several times compared to a decade ago while performance has significantly improved.

At the same time, some automakers are exploring the possibility of replacing LiDAR with pure vision solutions. Tesla is a key driver of pure vision solutions, leveraging deep learning algorithms to achieve environment perception based on cameras. This approach not only reduces hardware costs but also improves system computing efficiency. However, the performance of pure vision solutions in complex environments still needs to be improved, so most automakers still choose to find a balance between LiDAR and vision solutions.

4.2 Transformation of business models and software profitability

The promotion of intelligent driving technology has not only changed the technological route of the automotive industry but also driven the transformation of business models. Traditional car sales primarily rely on hardware revenue, while the introduction of intelligent driving functions provides automakers with new opportunities to profit from software. Automakers can continue to provide value-added services to users through subsequent function upgrades, service subscriptions, and other means.

For example, Tesla's FSD function adopts a subscription model, not only generating continuous revenue sources for automakers but also accelerating technology promotion. Other automakers have also followed suit, providing users with new functions and services through software updates. This "hardware embedding + software charging" business model not only enhances automakers' profitability but also promotes the rapid popularization of intelligent driving technology.

Conclusion

With the continuous penetration of intelligent automotive technology and the gradual promotion of vehicle-road coordination technology, the automotive industry is undergoing profound changes. The technological advancements in intelligent driving have brought users a safer and more convenient driving experience, while the introduction of vehicle-road coordination provides a strong guarantee for the comprehensive promotion of intelligent driving. In the future, with the continuous development of algorithms, computing power, and data technology, intelligent driving will gradually move from the laboratory to large-scale applications, bringing revolutionary changes to the global transportation system. At the same time, the industry also faces challenges in terms of technology costs, regulations and policies, and market acceptance. How to address these issues while rapidly developing will be the key to the future of the intelligent automotive industry.

-- END --