Entering 2023, people are still waiting for the first snow this winter, and at the same time lamenting that warm winter has become the new normal. But for the autonomous driving industry, this winter does not seem to be that warm.
Looking back on 2022, news of layoffs, bankruptcies, and departmental layoffs in the global autonomous driving industry are endless. The capital market has also gradually given up its investment craze for high-level autonomous driving since the beginning of 2021, and the overall market bubble has begun to burst. Some insiders believe that if capital continues to cool down, about 90% of autonomous driving projects will fail before large-scale commercialization.
But on the other hand, autonomous driving has indeed made great progress in technology and application, and has become a development direction recognized and determined by major enterprises, and even countries and regions around the world. Whether it is technology companies taking the route of Robotaxi and industrial cooperation, or new energy vehicle companies that regard autonomous driving technology as their main selling point, as well as established car companies that have embarked on the road of intelligence and electrification, they are all continuing to invest in the field of autonomous driving. In order to reach the dawn that is destined to come after the cold winter.
Under such circumstances, how to survive the winter for autonomous driving projects has become an issue that global car companies are paying attention to and exploring.
self-driving
Sufficient development momentum
But there are constant twists and turns
There is sufficient momentum for the development of autonomous driving, and society and the market's acceptance of autonomous driving is reaching new heights. In April 2022, the Ministry of Transport and the Ministry of Science and Technology publicly released the "14th Five-Year Plan for Science and Technology Innovation in the Transportation Field", which systematically planned the development direction of transportation science and technology represented by autonomous driving during the "14th Five-Year Plan" period. In November 2022, the Ministry of Industry and Information Technology issued the "Notice on Carrying out Pilot Work on the Access and Road Access of Intelligent Connected Vehicles (Draft for Comments)", announcing that it will conduct access management for L3 and L4 autonomous vehicles and launch pilot work.
With the raging industry push, no company will underestimate the long-term future of autonomous driving. But at present, the autonomous driving industry has encountered certain setbacks on a global scale. In 2022, Argo AI, the autonomous driving company invested by Ford, announced its failure. Aurora's stock price, which once attracted global attention, plummeted, and its market value shrank by more than 90%. The valuation of Mobileye, the industry leader, also shrunk from the previous US$50 billion to US$17 billion. Dollar. In the Chinese market, according to relevant data, the amount of financing in the autonomous driving field in 2022 will drop by about 60% compared to the same period in 2021. Behind the failure of travel projects, plummeting stock prices, and sluggish investment and financing is a deep-rooted and stubborn problem in the autonomous driving industry: cost inflation.
Autonomous driving for the winter,
The key is cost
So far, it seems difficult to find a car company that is not researching autonomous driving technology. However, the high cost of autonomous driving research and development has indeed become a key node that drags the autonomous driving project into climax. Car companies urgently need to ignite the fire of "cost reduction and efficiency improvement" on the road to autonomous driving research and development. Today, how to reduce research and development costs has become an issue that cannot be ignored in the field of autonomous driving.
Because high-level autonomous driving requires long-term development, and scenario-based market segments are difficult to quickly open up large-scale commercial space. Therefore, most autonomous driving projects are in the stage of long-term investment, but it is difficult to obtain a proportional commercial return. With the continuous investment in autonomous driving research and development, relevant data continues to accumulate and testing needs become more diverse, which in turn causes the overall cost of research and development to continue to rise.
However, the certainty of autonomous driving makes it difficult for the automotive industry represented by car companies to give up related projects. From this point of view, the key to easing the pressure on autonomous driving projects and helping car companies survive the short-term capital winter lies in how to reduce R&D costs.
Next, we might as well go deep into an autonomous driving project and explore from the specific technical processes and R&D links. Why does autonomous driving have such a huge R&D cost problem? How to solve these problems?
Peeling off the cocoon:
An autonomous driving project
Exploration of R&D costs
Our story is set in a European automobile manufacturing company. The company has a long history of manufacturing passenger cars and strong supply chain management capabilities. However, there is a lack of experience and foundation for digital and intelligent technologies such as AI algorithm development, massive data management, and autonomous driving testing. But at the current stage, autonomous driving has become a must-have option for this car company, otherwise it is likely to face the risk of being eliminated by the market in the next cycle.
Based on the estimates, the car company evaluated the research and development process of the autonomous driving project. This includes early-stage data work such as data collection, data import, and pre-processing annotation, as well as mid-term work such as autonomous driving algorithm training, simulation testing, and data archiving, and post-stage work such as algorithm deployment, real-road testing, and algorithm iteration. According to the car company's data volume prediction, the development cycle of such an autonomous driving project is about 60 months. In other words, companies must continue to invest for at least five years, and it is difficult to see commercial returns. After the actual implementation of the autonomous driving project, the car company quickly discovered a series of cost problems, represented by data storage costs, data usage costs, and technology research and development costs.
1. Cost pressure brought by data storage.
L4 level autonomous driving research and development will generate massive amounts of data every day, especially during the real-road testing phase. Each test vehicle will collect approximately 60TB of data every day. Just storing data for one car per day requires a hard drive expense. At the same time, the Corner Case obtained by refining these data is not "cold data" that can be stored inefficiently, but needs to be used frequently, so the efficiency of data retrieval is also a big challenge. The car company estimates that as its business advances, autonomous driving-related projects will collect more than 1,000 petabytes of data, which will need to be stored for at least 30 years. The data storage cost pressure this brings cannot be underestimated.
2. Comprehensive cost challenges arising from data circulation and application.
Autonomous driving R&D generates large amounts of unstructured data. For example, sensor data, camera data, lidar point cloud data, etc., and require a large amount of AI annotation training. The data infrastructure traditionally used by car companies is difficult to adapt to the needs of diverse data and AI development.
At the same time, as stated above. Autonomous driving development needs to go through multiple links such as data collection, import, preprocessing, training, simulation, and deployment. Autonomous driving data must flow between many process links. Each link requires different data types, load types, and access protocol types. A common practice in the industry is to store data in each link separately, so car company developers need to frequently copy and call data between different storage devices in different links, resulting in a lot of waste of valuable development time. In these processes, according to statistics from this car company, data copying and management time accounts for approximately 25% of the overall development time, which greatly slows down development efficiency. Although the costs caused by these data circulation, conversion energy, and applications may seem small, the accumulated pressure is astonishingly large.
3. Costs related to technical understanding and research and development have surged.
In addition to data problems, autonomous driving projects also face other problems. For example, the business process of autonomous driving is very complex and involves a variety of related technologies. Such as video encoding and decoding, radar point cloud, geographical information, vehicle sensors, vehicle controllers, AI, big data, OTA, etc., are all business categories that car companies have not covered before. These technologies require car companies to start R&D and exploration from scratch, which has actually resulted in a huge expansion of R&D costs.
The basic implementation method of autonomous driving is deep learning, which requires the use of the vehicle's perception + decision-making capabilities to achieve driverless vehicle driving. This has led to the development of AI algorithms becoming the core of autonomous driving research and development. Traditional car companies are not familiar with a large amount of AI development and training, nor are they familiar with the data, computing power, and network infrastructure required for AI development. They have to explore and build AI R&D systems on their own. This is also a major reason for the high cost of autonomous driving R&D.
Of course, in addition to these three points, autonomous driving projects still have many cost pressures, such as testing costs, AI computing power costs, etc. So far, we have learned about a car company’s autonomous driving research and development costs. If we want to help the autonomous driving industry overcome the cold winter and develop healthily, we must start from these specific issues to help car companies reduce costs and increase efficiency, and bring warmth and strength to winter.
Light up the fire of "cost reduction and efficiency improvement":
Huawei’s value on the road to autonomous driving
Overall, the road to autonomous driving for car companies has reached a crucial stage, and they need to overcome the winter that must be experienced. At this time, companies with technology and solution advantages need to contribute more of their own value to help car companies reduce the research and development costs of autonomous driving. In particular, it is necessary to prevent car companies from reinventing the wheel in large quantities, open up advanced and available industry commonalities, and reduce repeated development by car companies.
As mentioned above, the "cost inflation" of automakers' autonomous driving projects is divided into three levels: 1. The surge in infrastructure costs. 2. Technology research and development costs are too heavy. 3. The comprehensive costs of enterprises adapting to the new strategy of autonomous driving will increase.
If you want to solve the problem of "cost inflation" in autonomous driving projects, you can't just treat the headache and the foot. Instead, we need to comprehensively consider the needs of these levels to comprehensively promote the return of autonomous driving project costs to a reasonable level.
Under this multi-level, multi-objective need to reduce costs, Huawei's value advantage emerges. Huawei has an end-to-end autonomous driving solution that integrates decades of accumulated experience in network, storage, computing, AI, big data, and cloud. It can not only solve the problems of enterprises in digital infrastructure and technology acquisition, but also help car companies complete the overall intelligent transformation to adapt to the new stage and new goals, and ultimately bring comprehensive "cost reduction and increase" to car companies from three major levels. "Effectiveness" value helps car companies warm up this winter.
The first level is the significant reduction in digital infrastructure costs.
Take the data problem we talked about above as an example. Huawei has industry-leading storage technology. Its advanced multi-protocol interoperability technology can support full-process data storage with one storage solution, thereby avoiding frequent data copying. This just solves the R&D problems of autonomous driving. Car companies can use this to increase R&D efficiency by 25%, thereby accelerating the R&D process and reducing overall costs. At the same time, Huawei's ultra-high-density storage and automatic hierarchical processing technology for hot and cold data can effectively reduce TCO costs by 20%. At the same time, it supports large bandwidth and high OPS scenarios, which is more in line with the archiving, sorting, and recalling needs of massive data in autonomous driving research and development.
The second level is to help car companies obtain mature autonomous driving technologies and tools and try to avoid repeated development.
Huawei's autonomous driving development platform can bring a series of autonomous driving tools and platform value. For example, its exclusive cloud compliance collection and mapping can help car companies achieve flexible computing power scheduling and help autonomous driving projects complete high-precision map surveying and mapping work more conveniently and flexibly.
The massive data of autonomous driving may bring a lot of manual labeling and data cleaning problems to car companies. To this end, Huawei's autonomous driving development platform provides rapid and efficient cleaning and screening capabilities for massive data. Through automated data annotation, low-threshold and high-efficiency data processing is achieved, and the performance is 20% higher than the industry.
The shortest way for car companies to upgrade capabilities and services related to autonomous driving is to obtain relevant value from the cloud service platform. To this end, Huawei Cloud provides a series of empowerment and support focusing on autonomous driving capabilities. For example, AI algorithm development is the biggest technical challenge for car companies, with high threshold, long cycle and high uncertainty. To this end, Huawei Cloud provides a one-stop AI development platform, which can greatly reduce the difficulty and cost of platform construction, allowing car companies to focus on their core capabilities of algorithm development and model training.
The third level is to help car companies build an autonomous driving ecosystem and further complete digital transformation.
Huawei Cloud has joined hands with all sectors of the industry to build an ecologically open autonomous driving R&D platform, and has opened up Ploto, the open source code library for autonomous driving R&D platform solutions. It supports the deployment and docking of professional software service providers, and helps car companies quickly and accurately find relevant technologies and ecological cooperation. opportunity.
By integrating Huawei's technology and experience advantages, Huawei Cloud provides eight key capabilities for car companies, including: digital transformation experience, intelligent manufacturing capabilities, globalization experience, cloud-cloud collaboration capabilities, autonomous driving solutions, safety compliance solutions, and underlying technology Innovation, and open ecological cooperation capabilities. It can not only satisfy the "cost reduction and increase efficiency" of autonomous driving research and development of car companies, but also help car companies adapt to the development trend of digitalization and intelligence and improve industrial efficiency.
With the end-to-end autonomous driving capabilities provided by Huawei, we can see that a large amount of autonomous driving R&D costs are likely to be reduced and saved. Car companies can respond to challenges more calmly and efficiently, and are more determined that autonomous driving is the inevitable path.
The cold winter is about to pass, and everything is coming back to life. Humanity's first great revolution in the 21st century, the golden age of autonomous driving, is obviously coming soon.
