Home Tech From the darkest to the brightest, Tesla’s “behind the scenes” and “behind...

From the darkest to the brightest, Tesla’s “behind the scenes” and “behind the scenes”


Author / Luo Jing Edit / seal into On the one hand, it is surrounded by negative news, and on the other hand, Model Y’s sales are skyrocketing. Catfish Tesla’s every move always attracts the attention of the public. As a staunch supporter of renewable energy, Musk has spared no effort to promote electric vehicles to the public. But long before Tesla was born, there were electric cars in the automotive industry. In the 1920s, in economically developed cities such as New York in the United States, London in the United Kingdom, and Paris in France, more electric vehicles were driven on the streets than fuel vehicles. At that time, these vehicles used lead-acid batteries. But it was not until the advent of the internal combustion engine reversed the competition between fuel vehicles and electric vehicles, and the assembly line invented by Henry Ford made mass production possible. When one assembly line can produce 1 million cars, a car that originally cost $4,700 will drop to $380. Even ordinary blue-collar workers can afford a car. In 10 years, almost all electric cars on the street disappeared. History is always surprisingly similar. 100 years later, what happened to fuel vehicles is re-interpreted in electric vehicles, but the footnotes have undergone earth-shaking changes. Lithium-ion batteries, autonomous driving, on-board sensors, OTA…a series of new technologies are changing the definition of cars, and Tesla is undoubtedly the initiator and leader of this change. The Tesla model scares traditional car companies. The CEO of Volkswagen Group, the traditional car company, has publicly admitted that Tesla is leading the way in the field of autonomous driving software. Volkswagen needs to catch up. After all, no one wants to be the next Nokia. To this end, the world’s largest traditional car company has established a “Digital Car&Service” department, determined to become a software-driven company to meet the challenges of software-defined cars. And without the burden of fuel vehicles, Tesla has been in the electric vehicle industry since its inception in 2003, constantly pursuing mass production and reducing prices, and has developed a business model that is different from traditional car companies. With these two, Tesla can let more people buy cars, and the more users can buy cars, Tesla can use a large amount of user data to help optimize and iterate back-end autonomous driving algorithms and software Service, thus forming a positive feedback. Tesla is relying on sales, price, and technology to form a solid “iron triangle” to help Tesla, which is taking the route of pure visual algorithms, build a high wall that latecomers cannot overcome. 01 Price reduction is king Tesla loves to cut prices, it is obvious to all. The domestically produced M3 has been priced at 355,800 yuan since its release in May 2019, and has dropped 5 times in less than two years. Currently, on Tesla’s official website, the price of the lithium iron phosphate version of M3 + standard battery life + subsidies is 249,900 yuan, a decrease. Nearly 1/3 of the original. Another model, Model Y, also cut prices. At the beginning of the year, Tesla announced that the price of MY is about 150,000 lower than the imported version. Recently, it has been reported that Tesla will launch the Model Y equipped with lithium iron phosphate in July this year, and the price of Model Y is likely to fall further by then. Explore. Frequent price cuts have also made Tesla a frequent visitor to the headlines. The latter likes to call Tesla’s price cuts again and again as “cutting leeks.” But the actual situation is far from the conjecture of these media. Because since the day Musk took over Tesla, mass production and price reductions have been carved into Tesla’s bones. For the time being, back to 2006, Tesla, which has just completed its Series A financing, launched its first electric car, the Roadster, at the Santa Clara Auto Show in California. This is a niche and expensive two-seater roadster. At that time, Tesla’s weak financial resources made the company’s efforts to create two Roadster concept cars, one red and one black. As the only electric supercar in the audience, Roadster also lived up to expectations. , Attracted the attention of a large number of media and celebrities. At the press conference, Tesla announced that each Roadster car is priced at $90,000, and it can last 250 miles on a single charge. On the same day, 30 people promised to buy Roadster cars on the spot, including Google co-founders Brin and Pei. Odd, there are many billionaires in the field of science and technology. But man-made electric cars for the rich are only the first step of Tesla’s ambitions. Musk really wants to popularize clean energy and build an electric car that the public can afford. Only one month after the auto show, Musk proposed Tesla’s development roadmap “Master Plan”: 1. Build a sports car (Roadster); 2. Use the money earned to build a cheaper car (Model S/X); 3. Use the money earned to build a cheaper and best-selling model (Model 3); 4. While achieving the above, it also provides zero-emission power generation options. Automobiles are a typical technology and capital intensive industry, and countless start-ups have become cannon fodder. The fledgling Tesla is difficult to compare with traditional car companies in terms of manufacturing technology, supply chain management, and corporate branding. In addition, at that time, the entire electric vehicle industry chain was immature, the industrial supporting facilities were not perfect, and the battery cost was also high (10 times the current price). To be a popular and cheaper electric vehicle, both economically and practically It is very difficult to land in terms of operability. That being the case, it’s better to launch high-end electric sports cars for high-income groups from the very beginning, subverting people’s perception of electric vehicles with short cruising range and poor performance; when the time is right, they can reduce costs, mass produce, and promote cleanliness. Energy and other strategies. But even so, Roadster still encountered mass production problems, and the production cost exceeded the expectations of the Tesla team. Due to the lack of economies of scale in supply chain procurement, Roadster’s production costs have risen from an estimated US$65,000 to more than US$100,000, while Roadster’s price is only 90,000. In other words, every time a car is sold, Tesla will have to reverse the price. Ten thousand U.S. dollars. The heavy cost pressure put Tesla on the verge of a desperate cash flow break. The leaks happened in the night. In 2008, the global financial crisis broke out. Another Musk company, Space X, also failed to launch rockets one after another, unable to support Tesla. This electric car upstart unknowingly came to the brink of bankruptcy. Unexpectedly, the helping hand came from Tesla’s rival, the German auto giant-Daimler Group. In 2009, Musk received 50 million dollars from Daimler to renew his blood. At that time, Daimler Group was looking for a power battery provider for its Smart electric version. Musk heard that he directly bought a Smart and converted it into an electric version, and invited Daimler engineers to test drive. The electric car was amazed. After the contract was concluded, another US$50 million was invested in exchange for a 10% stake in Tesla. But this amount of money is really a drop in the bucket for Tesla, a “black hole”. At that time, Tesla’s annual revenue was only US$147.6 million, but business expenses were nearly 400 million, and a net loss of US$260 million. In addition, Tesla also needed a lot of funds to support the development and production of the second car Model S. The fire-fighting water still comes from the US federal government. After Tesla won the Daimler order and showed off the Model S concept car, the US federal government gave Tesla a low-interest loan of 465 million to encourage research and development of alternatives. Energy vehicles. In addition, Tesla also raised 226 million US dollars through the listing. After slowing down, Tesla won an order for an electric version of the RAV4 battery from Toyota, and at a low price of 42 million US dollars from Toyota CEO Akio Toyoda, Toyota and General Motors jointly built a car processing plant (now Fremont factory). It is reported that this factory is not only close to Tesla’s headquarters at that time, but also has an annual production capacity of 500,000 vehicles, laying a solid foundation for Tesla to control the cost of car production. The mass production and delivery of the second model Model S is on schedule. Compared with the first Roadster, the price of this four-door coupe is only 57,400 US dollars (standard endurance version), which is 63.7% of the Roadster, but it is not inferior in hardware performance, with the fastest acceleration of 4.4 per hundred kilometers. Seconds, the cruising range can reach up to 483 kilometers. More importantly, Model S creatively introduces a 17-inch central control touch screen, which integrates the entire vehicle’s information query, navigation, music, game and other functions on a single panel, which is like installing an iPAD for this type of car. , Users can enjoy the OTA over-the-air upgrade service of in-vehicle software on a regular basis just like enjoying the upgrade of a mobile phone upgrade system. This can ensure that each car owner can enjoy the latest software services, such as the Autopilot automatic driving assistance function launched by Tesla in 2014. Not surprisingly, ModelS sales far exceeded expectations, becoming Tesla’s first mass-produced car in the true sense. At the end of 2012, the pre-ordered Model S had reached 15,000; in 2013, Tesla delivered a total of 22,477 Model S. Compared with 2,500 Roadsters, Tesla’s mass production delivery capacity has increased by nearly 10 times. Relying on the great success of Model S, Tesla turned a profit in 2013Q1, but Model S is still a sports car, and the price is still high. The competitors it dropped are old luxury cars such as Mercedes-Benz S-series and BMW 7-series. The Model X that was launched in 2015 was not enough for civilians, and the price was around $50,000, so Musk also needed more affordable electric vehicles. It was not until March 2016 that Tesla launched the Model3, and Tesla ushered in a truly explosive model. Model3 Standard Edition starts at USD 35,000 and has a range of 354 kilometers, which is extremely cost-effective. Since its delivery at the end of 2017, Model 3 has continued to surpass traditional luxury fuel vehicle brands such as BMW, Mercedes-Benz, and Audi. At the same time, it has taken the first place in the market among similar new energy vehicle brands. According to statistics from EV Sales, from January to September 2020 The monthly global cumulative sales of new energy passenger vehicles was 1.784 million, of which Model 3 is the world’s best-selling model, with cumulative sales of 238,000, accounting for 13.4% of the world’s total. So far, the three-step strategy “MASTER PLAN” proposed by Musk in 2006 has basically been realized, but it is not enough. Comparing the prices of Tesla Model 3 and Model Y before subsidies and the selling prices of other top ten models in the US auto market, the price of Tesla cars is still 17%~31% higher. In other words, Tesla is still one step away from the price that the public can afford to buy a car. In February of this year, it was reported that Tesla would launch a Model series car with a price of 25,000 US dollars (160,000 RMB), which undoubtedly aimed directly at the current best-selling models on the market. On Tesla Battery Day 2020, Musk once said that starting with the heart battery of electric vehicles, he will reduce costs and increase efficiency in cell design, battery production, cathode materials, anode materials, PACK, etc., so as to reduce the cost per KWh. 56%, which in turn provides profit margins for price cuts. The Huachuang team calculated that after the price of Tesla Model 3 was cut by 23,900 yuan in October last year, it would need a sales increase of 20%-27% to obtain the previous total profit. Some investors also asked Musk at Tesla’s 2020Q3 investor conference, when the result of the price cut is that the company cannot achieve its profit margin target, will the company still cut prices? Musk said, “We hope that Tesla’s cars will be affordable to users, and we have to separate the concepts of affordability and cost-effectiveness. If the car is too expensive, consumers cannot afford it, regardless of It doesn’t matter what value proposition you have, so it’s important to be able to afford it…” It is not difficult to see that Musk is more concerned about the sales performance of the car than the price. And each price cut really stimulated Tesla’s sales. According to data from the Federation of Passenger Transport Associations, the domestically-made Model 3 lowered its price to 249,900 yuan from October last year, and its sales in October, November and December reached record highs, reaching 12,000, 22,000 and 24,000 respectively The domestic Model 3 2020 annual sales volume reached 135,000, in other words, Q4 single-season sales accounted for 42.96%. The same is true for Tesla’s full-year sales performance in 2020. Q4 sales exceeded 180,000 in a single quarter, accounting for 36% of the annual delivery of 499,647 vehicles. The side effects of the price cut are also obvious. The gross profit margin of Tesla’s automotive business in Q4 was only 24.1%, a decrease of 3.6 percentage points from the previous quarter, which was the lowest level in the year. The net profit attributable to the parent was US$270 million, which was in line with market expectations of 7.63. Billion dollars, far away. After the Q4 financial report was announced, Tesla’s stock price fell in response, and its market value shrank by nearly 60 billion U.S. dollars in one day. Since then, it has continued to fall. It has fallen to a March low of 539 U.S. dollars per share before stabilizing, and a total of 346 billion has evaporated. The trillion-dollar market value that once approached Apple has also become a bubble. Just when the public believed that Tesla was returning to a reasonable valuation range, many people believed that Tesla was undervalued. They said that Tesla will surpass Apple in the future to become the world’s largest company. The reason behind this is that Tesla is far from An ordinary car manufacturer, but a technology company with Saas service attributes. 02 The Meaning of Drunkard Selling a car is the face, and the software is the lining. For a long time, the automobile industry has relied on the manufacture and sales of new cars for profit, and automobile selling points have been focusing on the three major aspects of performance, appearance and brand. However, under the wave of electrification and intelligence, as a new dimension, in-vehicle software has become a new aspect of user purchase and profitability of auto companies. In other words, sales are transformed from purpose into means, and eventually flow to payment for in-vehicle software. When Tesla was the first company to run through this business model, the market’s perspective on this type of emerging car company changed. It was no longer the valuation of traditional automakers, but with SaaS service attributes. Valuation model for technology companies. Although the Tesla Model series will continue to reduce prices in 2020, Musk has also frequently increased the price of the on-board software FSD. Its Auto-Assisted Driving Software Package (FSD) has entered the price increase channel since January 2019, and has been gradually increased from the original US$6,000 to US$10,000 (or 64,000 yuan), which is equivalent to 1% of the price of the standard battery life version of the Model. /3, and the price increase momentum remains unabated. In order to inspire more car owners to use FSD, Musk changed the original one-time payment model and introduced a monthly subscription fee model ($100 per month). In the FSD that Tesla provides to car owners to purchase, it includes functions such as automatic assisted navigation driving, autonomous assisted lane change, autonomous parking, and smart calling. Among them, automatic driving technology is the top priority. In the “Key Points for Standardization of Intelligent Connected Vehicles in 2020” issued by the Ministry of Industry and Information Technology, autonomous driving technology is listed as a key element of the standard formulation and evaluation mechanism for intelligent connected vehicles. The research and development of autonomous driving is a typical artificial intelligence research and development engineering system, which is mainly based on three elements: algorithm, computing power and data. The algorithm solves the problem from zero to one, and the organization and optimization goals of the subsequent overall data engineering system. , Computing power is the physical basis for the realization of the model, and better algorithm efficiency and energy consumption indicators can be obtained through self-developed acceleration chips. But for autonomous driving, data is the core. Robin Li once proposed that the AI ​​era needs to switch to a new “AI thinking”. AI is different from the core of traditional software R&D system in that the implementation of traditional code functions is handed over to data and algorithms. Among them, the algorithm gap is easier to make up through open source and talent mergers and acquisitions, and data is the core barrier to differentiation. “Data spike algorithm “It is becoming the consensus of the industry. As the first electric vehicle brand to be equipped with an automated assisted driving system, Tesla also has the world’s largest assisted driving fleet, and has a huge first-mover advantage in terms of driving data. Beginning in 2019, sales of Tesla models have surged, and Tesla Autopilot’s mileage has ushered in explosive growth, from 1.73 billion kilometers in January 2019 to 4.828 billion kilometers in April 2020, doubling in a year, far exceeding Other competitors. The huge amount of real data makes Tesla’s data accumulation in high-precision maps and obstacle recognition significantly ahead of its competitors, and it is also more conducive to iterative algorithm updates. In contrast, other autonomous driving companies have been left behind. Behind. Musk himself announced on Twitter that the latest version of FSD Beta V9.0 has completed a new round of iteration. This time, the millimeter wave radar will be removed and become a complete pure vision solution, which is expected to be realized only by vision algorithms. Autonomous driving at L5 level. However, the new plan did not get the market’s appreciation. Instead, it attracted many doubts, believing that Tesla was too aggressive. Due to the low accuracy of camera sensors, there are some inherent shortcomings, such as poor night effects and susceptibility to extreme weather. Many experts believe that to achieve L4/L5 level of automatic driving, higher-end vehicle sensors such as lidar are needed. to realise. Lidar has high accuracy and detection range of up to 0-200m, which can greatly reduce the difficulty of software and hardware analysis, and can accurately predict objects at night. The day after Tesla’s official announcement of the withdrawal of the radar news, the National Highway Traffic Safety Administration (NHTSA) revised the Model 3 and Model active safety feature labels; after that, the authoritative third-party review magazine “Consumer Reports” It also announced the suspension of listing the 2021 Model 3 as a “recommended”. As public opinion continues to ferment, Musk has to explain to the public through the EV vertical media Electrek that even if there is no radar, the camera can achieve the work done by the radar to ensure the safety of autonomous driving. NHTSA is currently retesting the new model (without radar). In fact, Tesla is not the only company that takes the purely visual route for autonomous driving. The top players in the autonomous driving field, Mobileye and Baidu Apollo, both have products that are positioned on the purely visual route, SuperVision and Apollo Lite. Mobileye is an Israeli company. Tesla’s first-generation autopilot solution was handed over to Mobileye. However, after the death of autopilot in 2016, Tesla and Mobileye parted ways. But looking at the field of autonomous driving, Mobileye is a recognized top supplier in the industry. Its EyeQ series of chips have been adopted by 27 automakers, many of which include traditional car giants such as General Motors, Volkswagen, and BMW, as well as other car manufacturers such as Weilai and Ideal. New forces. Just last September, Mobileye 360° pure vision sensor solution SuperVision debuted at the Beijing Auto Show for the first time, serving the Geely Group’s high-end car brand-Lynk & Co Electric Vehicle Zero. It is reported that the solution is equipped with two EyeQ5 chips with a 7nm process and a computing power of 24TOPS. It is equipped with seven long-range cameras and four parking cameras. There are no sensors such as lidar and millimeter wave radar. Baidu, which has been using lidar, also launched its own pure-vision autonomous driving solution Apollo Lite at CVPR in 2019. From the multi-sensor solution to pure vision, the sudden change of Baidu made the public feel puzzled. At the meeting, Wang Liang, chief of Baidu’s Autopilot Committee, said: “Many sensor fusion scheme designs are more complicated, and technicians often design algorithms from the perspective of quickly solving problems. In this process, it is inevitable to use the respective advantages of heterogeneous data to complement each other to avoid difficult problems.” “Although the multi-sensor fusion scheme designed based on this idea can circumvent the problems that the single-sensor scheme is difficult to solve in the short term, in the long run, the design of deep coupling between data and strategy is not conducive to providing real redundancy for the environmental perception system.” “In the traditional strategy of lidar-based and vision-assisted, the problems and deficiencies of visual perception itself are not sufficiently exposed under the cover of radar perception. Therefore, the problem of visual perception needs to be independent to be better solved.” At the same time, Wang Liang also expressed his views on the camera: The camera is a relatively mature sensor. In addition to the advantages of light weight, low cost and compliance with vehicle regulations, the development trend of imaging technology with high resolution and high frame rate (imaging frequency) means that the image contains more environmental information, and the video data is also It is most similar to the real world perceived by the human eye, but compared with three-dimensional point cloud data, the information in two-dimensional images is more difficult to mine, requiring the design of more powerful algorithms, the accumulation of large amounts of data, and longer-term R&D investment. A brief summary of the above is that there are potential risks in taking the multi-sensor fusion route. The pure visual route can make up for this defect, but the pure visual route requires a large amount of data accumulation and long-term research and development to train the background algorithm, and data acquisition and algorithm deduction are extremely barriers. It is difficult for most car companies to adopt in the short term. It can be seen that the pure visual route is the kind of difficult but correct thing that everyone often says. At present, Mobileye and Apollo have both pure vision solutions and multi-sensor solutions such as lidar. Tesla has chosen a way to go to the dark. One example is the removal of millimeter wave radar. The “courage” behind Tesla is From: self-developed chip FSD, self-supervised learning neural network algorithm, supercomputer Dojo. Someone once helped Tesla calculate an account. At present, there are more than 820,000 Teslas equipped with HW 2/3 hardware driving around the world. The average user drives about an hour per day (8 cameras per car). ), the fleet generates approximately 196.8 million hours of video every month. The huge amount of data puts high demands on the computing power of autonomous driving chips. Before self-developed FSD, Tesla had used the Mobileye EyeQ3 chip and the NVIDIA Drive PX2 chip, but their computing power could not meet Tesla’s needs. Therefore, in 2015, Musk personally found Jim Keller, the chip god who just resigned from AMD, and asked him to develop Tesla’s self-driving chip. The importance of self-developed chips is to achieve deep adaptation of software and hardware to achieve minimum energy consumption and optimal performance. Among smart phone brands, Apple, Huawei, and Samsung all do this. The biggest feature of FSD’s self-developed chips is that they “only serve one customer-Tesla”, while chip companies such as Mobileye and Nvidia have many downstream customers and have to make a general chip solution. At present, the computing power of the FSD platform is 144TOPS, the computing power of a single NPU reaches 72TOPS, the preparation process is 14nm, and it is equipped with HW3.0. But just in Q4 of 2020, TSMC said it will take Tesla’s 7nm order and use the latest InFO_SoW packaging technology. At that time, the computing power of FSD is expected to increase by 4 times, and the probability of increasing to more than 500 TOPS is very high. In addition to the chip, Tesla trains a set of neural network algorithms that can be self-supervised learning. As of April 2020, Tesla’s true road test mileage has accumulated to 4.8 billion miles. If these data are all manually labeled for algorithm iteration, it will undoubtedly be a huge expense. To this end, Tesla’s senior AI director Andrej Karpathy launched the “Operation Holiday” data automation plan. Tesla’s deep neural network will directly mine the characteristics of the data itself, instead of manually labeling the data set, so as to carry out the perception algorithm. Continuous iteration, the ultimate goal is to complete full automation, without human intervention throughout the process. Just when everyone was wondering why Tesla supports huge data for deep learning, the cloud training supercomputer Dojo behind it also surfaced. Generally speaking, supercomputers are national-level special equipment, providing super-fast computing speed and super-large storage capacity for the high-tech fields and cutting-edge technology research needs of various countries. Among the top 10 supercomputers in the world in 2020, Japan’s Fujitsu topped the list with a peak performance of 0.415 exaflops; the second place in the United States was Summit with a peak performance of 0.148 exaflops; China’s Shenwei Taihu Light ranked first Fourth, the performance is 0.093015 exaflops. How does Tesla Supercomputing Dojo perform? Musk once said that the peak performance of the Dojo supercomputer should reach the level of exaFLOP per second, which is a trillion floating point operations. (Original words: A truly useful exaflop at de facto FP32.) At that time, Dojo will surpass Fujitsu of Japan and become the world’s number one supercomputer. These bold remarks aroused countless speculations by the public, and at the 2021 Computer Vision and Pattern Recognition Conference on Monday, Karpathy disclosed some of the operation of Tesla’s latest version of Supercomputing, which will serve as the development prototype of Dojo. . It is reported that the current Tesla supercomputing operating speed is 1.6Tbps per second, and the peak speed per second is 1.8Eflops. Karpathy said: “If you count FLOPS, this supercomputing can rank fifth in the global supercomputing rankings. This position is currently occupied by Nvidia Supercomputing Selene. This cluster has a similar architecture to Dojo and a similar number of GPUs (4480 vs. 5760 for Dojo).” This supercomputer has helped Tesla process more than 1 million videos of about 10 seconds, and marked the distance, acceleration and speed information of 6 billion objects in the video. The data storage space has reached 1.5PB. This new version of supercomputing is not the ultimate version of Dojo. Tesla’s supercomputing is still iterating and evolving. Musk’s goal of computing power to reach 1exaflops of Dojo is expected to be unveiled on Tesla AI Day at the end of July this year. . 03 The future is here Tesla’s various layouts for autonomous driving have long gone beyond the scope of traditional car business, and subverted the latter’s perception of cars. In 2018, the American authority “Consumer Report” pointed out that Tesla Model 3 has the problem of too long braking distance, so it did not recommend it. Tesla engineers upgraded the system through OTA (Over-the-Air Technology), shortening the braking distance by 6 meters, and solved the problem in just a few days. It is almost unimaginable that the replacement takes place in traditional car companies, because the solution of traditional car companies is most likely to be through large-scale recalls, or replacement of parts through 4S stores, which requires car owners to wait a long time. OTA is not a new thing. It refers to remote application management of mobile terminal devices through the network. Most people can experience it on their smartphones, such as mobile phone system upgrades and application software updates. Analogous to mobile phones, car OTA can also be divided into two categories: FOTA (Firmware OTA) firmware update and SOTA (Software OTA) software update. Most traditional car companies can implement SOTA, which can remotely update car map display, audio, video and other infotainment systems, but like Tesla, it can’t do anything to update the brake system. The latter involves rewriting the firmware program to realize the car. Renewal of power system and chassis system. Behind this can and can’t, reflect the deeper reasons, the difference between the electrical and electronic architecture (Electrical/Electronic Architecture) at the bottom of the two. Cars have many functions, ranging from the control of lighting, wipers, and windows to the ABS (brake) system, assisted driving system, map display, audio-visual player, etc., all handed over to Electronic Control Units (ECUs) to realise. The electrical and electronic architecture is the layout of these ECUs, which can be roughly divided into three categories: distributed, domain centralized, and centralized. Under the traditional distributed EEA architecture, each ECU has a corresponding relationship with the realized function. The hardware and software functions are deeply coupled. When the car has more functions, it also means the number of ECUs and wiring harnesses, which greatly increase the weight and cost of the car body. . These ECUs come from different suppliers and have different embedded software and underlying codes. The computing power cannot be coordinated, and the OEM has no right to maintain and update the ECU. In addition, different ECUs may have overlapping functions, leading to mutual redundancy and increasing unnecessary costs. In this context, traditional cars want to update their technology or firmware and rely more on the iterative completion of the model. That is to say, consumers need to keep changing new cars to enjoy the latest technical services, especially when it comes to chassis and power system updates. Distributed architecture is undoubtedly a big shackle for the intelligent upgrade of automobiles. In this regard, for the first time, Tesla adopted a centralized electronic and electrical architecture, using the central processing unit to uniformly manage different domain processors and ECUs, decoupling hardware and software, and opening up all hardware resources, including firmware and Complete vehicle OTA including software. Under the framework of centralized electronic appliances, automobiles are moving towards wheeled computers. Su Jing, head of Huawei’s autonomous driving product line, once said: “Traditional automakers think that the base of a car is a car, and try to embed a computer on the basis of the car. We think that the basis of the car is a computer, and the car is a computer-controlled exterior. Assume.” As early as June 2012, Tesla carried out an OTA upgrade to the first production car Model S. In 2017, Tesla implemented a centralized EEA architecture for the first time in Model 3 models. Since then, Tesla has carried out more than 120 OTA updates, including software OTA for autonomous driving, entertainment applications, and batteries. Firmware OTA for management and improvement of brake horsepower. In 2019, the foreign automotive media Top Gear released a performance test comparison video between Porsche’s Taycan Turbo S and Tesla Model S Performance in the tubing. The Taycan Turbo S accelerated at 402m, 0-96km/h, and 0-161km/h. Win the test. Some netizens questioned the rationality of the test, saying that the model tested by Top Gear is not Tesla’s latest Model S, but an old model from 2017, and the newly released 2019 new model has a significantly improved drive system and endurance. . Musk replied to the comment, agreeing, and let Tesla engineers optimize the motor control of Model S, and increase the peak output power of Model S by 50 horsepower through OTA. The upgraded Model S will be able to ensure that it is 402m. , 0-96km/h, 0-161km/h acceleration test, beat the Porsche Taycan with a greater advantage. Smart cars represented by Tesla are reshaping the definition of “cars” in the automotive industry. In the past, the completion of the delivery of a new car has meant the beginning of depreciation. The biggest change that Tesla has brought is to continuously improve the performance of the car through frequent use of data collected from the OTA (Over the Air) interface. In the process of using the car, a large amount of data and demand will be generated, which will be fed back to the car companies. On this basis, the car companies will continue to optimize and iterate their products, and then synchronize them to users in the form of vehicle OTA to continuously improve user experience. This is very similar to smart phones, and it also gives Tesla the same characteristics and core advantages as Silicon Valley technology companies. Through mass production and price cuts, Tesla continues to attract more car owners into its self-built ecosystem. The large amount of data in the user’s use process helps Tesla deep learning and continuous improvement of products. It is updated and upgraded through vehicle OTA. Software and hardware, adding new ones, replenishing old ones, and constantly improving. end Although the performance of new energy vehicles in the policy and investment market is turbulent, the actual situation is that new energy vehicles have not yet reached the stage of rapid growth. In 2020, the global penetration rate of new energy vehicles is less than 5%, and the domestic penetration rate is 5.9%. There are many factors that hinder the high volume of new energy vehicles, such as battery safety, technological level, and completeness of supporting facilities such as charging, and another is price. In China’s 100,000-200,000-price range car market, the penetration rate of new energy vehicles is obviously not as good as other ranges, but this group accounts for 43.2% of the total. This part of consumers is more about the extreme price-performance ratio of the car, and the current electric car The cost of the power battery is still relatively high, so the price is significantly higher than that of fuel vehicles of the same level. On Tesla’s Battery Day in 2020, Musk announced to the audience that he would cut the cost of battery production in half, and plans to launch models priced at about US$25,000 (about 160,000 yuan) in the next three years. By then, it remains to be seen whether the catfish Tesla can wave its tail again and stir the market. references: books Ashley Vance “Iron Man of Silicon Valley: The Adventure Life of Elon Musk” Hamish Mackenzie “Tesla Biography: Realizing the Impossibility” Research report Guojin Securities “160,000 “Model 2″: Realization Path and Industrial Chain Impact” Guosen Securities “Discussing Tes Based on the Perspective of Business Model Reform” Tianfeng Securities “Tesla: From monthly data charges to SAAS-like business model, subverting the traditional auto industry” Industrial Securities “It is an electric car, but also a smart car. Tesla series report (4) detailed automatic driving system” Open Source Securities “Domain Controller-The “Brain” of Smart Cars” Zhongtai Securities: “Intelligent Driving Topic: Discussion on the Commercialization Path, Industrial Evolution and Investment Opportunities of Automobile Intelligentization” News Quasar Channel: Dojo super counts, Tesla challenges the visual limit Quasar Channel: Why Tesla “ditched” millimeter wave radar Quasar Channel: Baidu Mobileye Tesla, the “revival” of the pure visual route of autonomous driving? EV Vision: Model S will get another 50 horsepower performance improvement through OTA CICC’s finishing touch: new trends in automotive electrical architecture, in-vehicle communication welcomes changes