In 2016, McKinsey released a report of great concern:
《Advanceddriver-assistan cesystems:Challengesandopportunitiesahead 》
(authors: seunghyuk Choi, Florian thalmayr, Dominik wee, and Florian weig).
Thanks to McKinsey and the original author for their efforts and work for the healthy development of advanced driving assistance system (ADAS). Link to the original text: http://www.mckinsey.com/industries/semiconductors/our-insights/advanced-driver-assistance-systems-challenges-and-opportunities-ahead .
This report collects and analyzes a large number of information resources about the future development trend of ADAS, which provides valuable reference information for ADAS suppliers and investors; visspect interprets the important information in this report, and is willing to share and discuss with peers.
1、 The camera solution should be the main opportunity for semiconductor manufacturers in ADAS industry
McKinsey: We surveyed a large number of semiconductor manufacturers and analyzed their opportunities in ADAS industry before 2025. Various factors were taken into account in the survey, including forecast of end market adoption, system, component price decline, etc.
The conclusion of the analysis is that the total revenue of semiconductor manufacturers in ADAS industry is expected to grow steadily, reaching about 4.6 billion to 5.3 billion US dollars in 2025. Breakdown of total revenue:
1. For the system: “Parking Aid System” is expected to bring the most revenue for semiconductor manufacturers. Next are “automatic emergency braking”, “adaptive cruise control” and “forward collision warning (FCW)”.
2. For system components: as shown in the following figure “revenue forecast of semiconductor manufacturers in ADAS industry”, it is estimated that the revenue growth points are processors (accounting for 37% of the total revenue) and sensors (accounting for 28% of the total revenue).Interpretation of McKinsey report: which direction of ADAS will become the mainstream?
Figure: Revenue Forecast of semiconductor manufacturers in ADAS industry
Interpretation of leisureauto: according to ADAS industry experience, the cost ratio of processor to sensor is about 1:1. According to the above figure “income forecast table of semiconductor manufacturers in ADAS industry”, it is estimated that the price range of millimeter wave radar manufacturers expect in the future is (double reference and single reference schemes, which are the lower limit and upper limit respectively): ”
The lower limit of leisureauto calculation: lens module cost X (13% / 28%) = RMB 100
Upper limit of leisureauto calculation: lens module cost X 13% / (37% – 28%) = RMB 350
For leisureauto, the product can be completely combined with radar technology. The basis of cooperation or merger with radar team is that after the radar technology is stable and a formal industry is formed, the price will be greatly reduced. According to the current market situation and trend analysis, it is not feasible to reduce the price of millimeter wave radar to the lower limit of RMB 100 / set.
Therefore, leisureauto can be reversed: under the influence of the price factor, the possibility of implementing the dual authentication scheme is very low, and the camera scheme will become the mainstream of the ADAS industry.
General purpose chips will become the mainstream, and the emerging ADAS team will gain new opportunities
McKinsey: it is estimated that processors (37%) and sensors (28%) will account for the majority of the total revenue of semiconductor manufacturers. Semiconductor manufacturers will consider “differentiation” to enter the market segments. They can gain value by strengthening the system and expanding their products to “software” and “algorithm”.
Therefore, manufacturers with experience in consumer electronics, mobile devices or data processing can most effectively improve processor performance, and these performance enhancements will become the most important selling points. Because fast processors are built on the basis of miniaturization technology, manufacturers need to invest a lot in R & D and manufacturing. The sales revenue of manufacturers only in the automotive market can not fully meet the return on investment, and they also need other revenue sources to obtain better return on investment.
In addition, manufacturers with experience in neighboring industries can adapt to ADAS applications faster by changing the performance of some products embedded in ADAS, thus greatly reducing the development time. For example, NVIDIA refitted the Tegra platform originally developed for game devices, smartphones and tablets, and soon used it in automotive systems.
Leisureauto interpretation: the comparison between NVIDIA’s route and Mobileye’s original route, that is, the comparison between general chips used in special industries and special chips designed by special industries, the former should become the mainstream. Based on this, leisureauto speculates that Mobileye may be forced to transform its technical solutions, and even lead to some problems; Neusoft in China may get a lot of opportunities in the next two years. Leisureauto and other teams with formal products will strengthen market development in the next step.
3、 Algorithm will not become the threshold of ADAS in the future, and the prospect of product orientation of ADAS industry
McKinsey: In addition to the processor, the sensor performance will also be greatly enhanced. At present, there are many different types of sensors in the market. For ADAS, the following three types of sensors are the most important:
The first type, the most cost-effective type of optical sensor, is that the optical sensor is generally based on the camera scheme. This type of sensor is multi-functional and can realize a variety of ADAS functions, but they are vulnerable to adverse weather conditions and other environmental hazards. In addition to using cameras to collect the image data of vehicles and pedestrians around the carriageway, complex software algorithms are also needed to process, recognize and analyze the image data.
The second is lidar system, which uses laser scanning to generate 3D images of the surrounding environment. Compared with optical sensors, lidar can detect the location of objects directly, which is not sensitive to the surrounding weather conditions. Lidar has a large detection range, which can reach about 100 meters around the vehicle. It usually needs external installation. In the past ten years, although the price has dropped from tens of thousands of dollars to thousands of dollars, the cost is still very expensive. To popularize the use, we must further reduce the price. In addition, the short-range lidar system can detect a distance of about 10 meters in a single direction. This kind of lidar has been used in some cars, such as the short-range lidar provided by continental group.
The third type, electromagnetic wave radar, is used to determine the distance, velocity and direction of an object in short and long range. ADAS usually uses the radar signal in the range of 20 to 80 GHz. These sensors are better than other sensors in adverse weather conditions, but the measurement range and angle will be weak. For example, the electric Doppler radar can detect obstacles up to 250 meters high. Because the measurement angle is quite narrow, the adaptive cruise control usually combines the electric Doppler radar with the short-range lidar to make it have a wider measurement angle.
Although the revenue generated by sensors (accounting for 28% of the total revenue) is lower than that generated by processors (accounting for 37% of the total revenue), semiconductor manufacturers are more willing to work hard and invest in the field of sensors, because this field is easier to be subdivided, and the requirements for enterprise size are not very strict. In fact, different sensors have different characteristics, according to their different performance levels and signal processing capabilities, they are used in ADAS products. Just like processors, if semiconductor manufacturers have experience in adjacent industries, they can quickly gain advantages. For example, semiconductor image sensors created for the consumer market can be quickly expanded to ADAS sensors to meet the safety requirements of automobiles and the specific requirements of automobiles.
In addition, it will be very difficult for semiconductor manufacturers to enter the software field, because they lack advanced software technology foundation. In order to compete, they need to establish their own software teams, or merge and acquire external software teams.
To sum up, semiconductor manufacturers integrate hardware, non silicon products (including drivers, operating system adoption and codecs) and algorithms (including software for real-time processing of sensor data), such as improving memory while improving the performance of central processing unit. This integration enhances hardware and generates more value. In addition, semiconductor manufacturers will try to provide more modules or integrated solutions, such as SOC. ODM vendors tend to prefer single component solutions, such as modules and integration solutions (including software), because they provide better performance, lower development costs, and lower dependence on first tier suppliers.
Leisureauto interpretation: this part of the content implies that the liquidity of the “algorithm” itself is likely to decline significantly in the future and become the standard library of the chip. Once the “algorithm” at the application level is mature, it will spread quickly, and its value will fall precipitously; the “algorithm” for low-frequency special applications still has its own living space. At present, some chip or ADAS module suppliers provide functions, which are almost free of charge, but the test results are different. For example, leisureauto’s M1 module adopts the latest “algorithm” developed by itself, and the consistency rate is very high when benchmarking with Mobileye; however, the Korean module is quite different from Mobileye in the benchmarking process.
The word “algorithm” itself is an inaccurate term. The correct term is application software. Algorithm refers to mathematical method in particular. Its basic method has been formed in the 18th century. Various application methods in CV field are the achievements of mathematical circles, which are shared by all mankind. No one can and may monopolize them.
In the production process of ADAS, the essence of constantly adjusting various parameters and choosing various methods in practice is an engineering problem, which can be solved by money. Especially in recent years, with the introduction of various AI tools, the development has become faster, and even a large number of developers no longer need to have a strong mathematical foundation, because the mathematical part of ADAS has been replaced by “development tools” The group solved the problem, and the results were provided to the developers of ADAS software in a nearly free way.
At present, the effect of leisureauto’s own “algorithm” is very good. On April 1, 2017, it surpassed the recognition distance of Mobileye for the first time in the product test. Although there are some accidental factors, it always reflects that the domestic team is not unable to meet the standard of foreign big manufacturers in software.
To sum up, “algorithm” will not become the threshold of ADAS in the long run. If you are determined to become a great enterprise, in addition to technical ability, the vision and ambition of the enterprise can be transformed into the dream and business philosophy of concrete action; outstanding and visionary entrepreneurship and management team; huge market space, good business model, clear and feasible enterprise development strategy; extraordinary competitive advantage, excellent management and implementation Strong, innovative products, changing from time to time, and knowing how to lock in and interact with customers are the things leisureauto can rely on.
If you are looking for reliable OEM ADAS manufacturer, please feel free to email [email protected], we will offer professional & customized ASAS solution at reasonable price.