This is the 1609th push of Yangtze River Chuangchuang,
At the end of the article, click "Watching" to let me know that you are watching.
Autonomous driving is a veritable hot track in 2021. In addition to the layout of giants such as Tencent and Huawei, the financing amount of the autonomous driving track in 2021 is as high as 40.74 billion yuan, which is the hottest market among all the tracks in the automotive travel industry.
In recent years, both foreign companies and domestic companies have carried out deeper exploration and research on automatic driving, but the problem of car manufacturing is still a headache for head technology companies. Compared with car building, Musk seems to be more confident in Tesla's AI capabilities, and the most important and valuable application area of AI is undoubtedly autonomous driving.
On March 6, 2022, Tesla CEO Musk sent a Weibo post: "Even some of the world's best artificial intelligence software engineers have not realized the advanced degree of Tesla's artificial intelligence." ”
Will the future of autonomous driving be a car or a robot, and as autonomous driving evolves to a higher level, will the core of the full-stack capability of autonomous driving still be computing power? This issue will analyze and answer these questions for you.
Source | Digital krypton
Author | Ben
Edit | Shi Yaqiong
March 6, 2022, 4:00 p.m. Beijing time
Tesla CEO Musk sent a tweet
Previously, on January 27, 2022, Musk said on a conference call: The development of humanoid robots will be the most important work this year. As soon as this statement came out, the industry was in an uproar, and various analyses and speculations followed.
At the Tesla AIDay on August 20, 2021, Musk announced the Tesla robot TeslaBot, which looks less reliable, and the TeslaBot on the scene is dressed up by a real person.
AI Day has always been Tesla's big day, when TeslaBot overshadowed FSD, neural network autopilot training, D1 chips, and Dojo supercomputers.
In the future, TeslaBot will use the D1 chip with the horror computing power, the Dojo supercomputer, and Tesla's proud visual neural network, etc., that is, TeslaBot and FSD use the same full set of computing equipment.
Musk also said that Tesla will be the world's largest robotics company, and "our cars are like semi-sentient robots with wheels."
In other words, autonomous driving is not only an accessory to the car, but also the computing power, algorithms, perception, execution, and even data related to automatic driving can become the core capabilities of AI to open up more vertical fields.
This view, I believe that not only Xiaomi and Baidu, which have released robot products in the same month as Tesla, will nod their heads and say yes, but also many technology giants will agree. As early as 2018, the most advanced technology companies in the United States, except for Facebook, systematically laid out automatic driving.
The autonomous driving layout of the US technology giant
Although it has not yet been supported by scientifically calculated data, in the future, as high-level automatic driving technology continues to break through the scene limitations, MaaS has landed on a large scale, and the efficiency of passenger cars and commercial vehicles has been greatly improved. The total number and increment of the car will certainly show a downward trend, and even the shape of the car may change subversively.
Tesla is the only technology company that has achieved full-stack self-development and self-production in the core areas of new energy vehicles and autonomous driving technology, and what it does is of great reference value. Xiaomi and Apple, which have been entangled in the series of issues of car building and software-defined cars for many years, and have also fought on the mobile phone battlefield for many years, are of course worthy of attention in the field of automatic driving.
Although Tesla's mobile phone may not be serious, Xiaomi's car must be serious. However, before the real car is built, Xiaomi has first made a systematic investment layout in the field of automatic driving.
Early on, in 2014, Xiaomi invested in Kailide;
In 2017 and 2021, Shunwei successively participated in Momenta's A+ round and C round of financing; in 2017 and 2018, Shunwei invested in The Wisdom Walker A and B rounds; in 2018, Shunwei strategically invested in Youdian Technology; in 2019, Shunwei invested in Beixing Photon and Aoix; in 2020, Xiaomi contacted and invested in The Network of Cards, BYD Semiconductor, and Energy Chain Group.
With the official announcement of the car in March 2021, Xiaomi invested in 6 autonomous driving companies from June to September, including: ADAS supplier Zhongmu Technology; Lidar intelligent manufacturer Hesai Technology; 4D millimeter wave radar supplier Geometric Partners; smart parking solution Shangai Parking; high-precision positioning solution provider DeepMotion; and autonomous driving chip company Black Sesame Intelligence.
Soldiers and horses have not moved, grain and grass have gone first, Xiaomi has not yet built a car, but first bet on automatic driving and related companies in batches, and people can't help but wonder why automatic driving is so important for car manufacturing? What is the purpose of car building at this stage? What will the automotive market look like in five or ten years?
On the other side of the ocean, the three trillion giant Apple has also struggled for many years on the issue of car building.
As early as 2014, Apple began a plan code-named "Titan", quickly assembling hundreds of employees to participate in the secret development of Apple cars. The project was approved by CEO Tim Cook, and at the time, a large number of analysts reported that Apple's decision was intended to counter Tesla.
Unlike Apple's previous main consumer electronics products, the supply chain system of automobiles is hundreds of times that of products such as mobile phones, and in 2016, due to the resistance encountered in the construction of the vehicle manufacturing industry chain far beyond imagination, Apple suspended the "Titan" project and turned to the development of autonomous driving technology.
By June 2021, according to foreign media reports, Apple has restarted the "Titan" plan, not only to build the whole vehicle, but also to develop automatic driving software.
From 2020, or earlier, it was rumored that Apple has a good feeling for the new car-making force in the United States, and it may be biased to define Canoo as a new car-making force, because Canoo does not intend to sell cars, and its core is a pure electric development platform that integrates power, power, perception, execution, and computing power in the skateboard chassis, based on which to adapt to different functional bodywork. It is said that this kind of car can shorten the development cycle by 6-12 months and reduce the development cost of the whole vehicle by 60%.
Canoo's LEGO-style car.com
In the end, apple's hoped deal did not work out, but it was recently reported that several of Canoo's executives, including CEO Ulrich Kranz, left and joined Apple after leaving.
Apple has always been good at deconstructing products that seem to be mature, subverting the original market from the perspective of customer experience, for mobile phones and wearable devices with only a history of more than ten years, Apple has gained a fairly high success rate, but it is difficult for cars with a century of history, starting from automatic driving, using software to redefine cars is undoubtedly a shortcut that is not simple.
A senior investor in the field of hard technology said that the automobile industry stands at the commanding heights of the civil industry, and automatic driving stands on the shoulders of giants. The relationship between the future car and automatic driving, just like the smart phone and the operating system, there are still many variables, just like in addition to iOS and Android, there have been many other intelligent operating systems, but they are gradually dying out; and Google's Android system, although successful, but his mobile phone is not selling well. The complexity of the car is far more than the mobile phone, and the automatic driving integrates the most cutting-edge technologies such as AI, and the relationship between the car and the automatic driving is also more possible.
Cars or robots
Autonomous driving may be breaking through the extension of transportation
On August 18, 2021, at the Baidu World Congress, Robin Li released the Apollo "automotive robot", and at the same time, the robot concept car debuted. On the outside, the automatic gull wing door, all-glass roof and external sensors are integrated; inside, the automotive robot does not have a steering wheel, pedals, and has intelligent configurations such as oversized curved screens, intelligent consoles, variable glass, and zero-gravity seats.
A few days earlier, on August 10, the eleventh anniversary of Xiaomi, Lei Jun gave an annual speech with the theme of "My Dream, My Choice". In addition to the release of mobile phones, tablets and other products, it also launched a blockbuster new product, Xiaomi's first bionic four-legged robot "CyberDog", and affectionately called it "Iron Egg".
Coincidentally, a few days later, on August 20, Musk announced at an artificial intelligence event held by Tesla's ministry in Palo Alto that Tesla will launch intelligent robots in 2022.
TeslaBot
Musk said the Tesla robot is a simulated robot that shares a car AI system, and the supercomputer can both train cars to navigate city roads and train robots to do the same thing, and this robot will be able to complete tasks including going to stores, organizing household items and other daily tasks.
Within half a month, the old self-driving companies, new car-making forces, and the world's highest market capitalization OEMs are all releasing robots, and even if they are accidental, there must be deep internal connections.
Recently, Service Robotics, a robotics division spun off from Postmates, launched a new generation of delivery robots with functions such as automatic emergency braking, vehicle collision avoidance and fail-safe redundancy of mechanical braking, trying to create a new category, that is, robots with L4 level of autonomous driving.
The robot's AI computing platform, provided by Nvidia and equipped with Ouster's 3D lidar sensors, has successfully completed tens of thousands of deliveries in Los Angeles in 2021.
Uber's delivery robot
It is not easy to distinguish between cars and robots by precise definition, but if you return to the essence of business, the difference is obvious, in fact, the robot here omits the definite word, that is, its essence is a commercial robot, the car can be understood as the end consumer will eventually be sold to the end consumer, and the robot is a service that needs to be subscribed to in MaaS.
Standing on the high ground of artificial intelligence in the future, the margins of car, RoboTaxi and Robot will become increasingly blurred. In the near future, whether to sell cars, or do operations, or do all of them, will also be a choice problem that traditional car companies, the current new car-making forces, and autonomous driving upstarts will face.
Since 2021, roboTaxi's costs have been rapidly declining. According to public information, Baidu's fifth-generation unmanned vehicle Apollo Moon, the cost of a bicycle is 480,000, which is roughly half of the cost of Robotaxi in the previous year, using THE αT of BAIC ARCFOX Jihu as the foundation, integrating the Apollo automatic driving kit, including 2 lidars, one in the roof of the car, the other under the front license plate, as a system redundancy, there are a total of 13 cameras around the body, 5 millimeter-wave radar, with Apollo cloud, Up to 800 TOPS hash rate.
In the interview, Yuanrong Qixing said that in terms of technology, Yuanrong Qixing first proposed self-developed inference engine, and multi-sensor fusion, and adopted an integrated sensing solution, which has reduced the cost of L4 level automatic driving hardware to 250,000. In addition, its front-loading self-driving solution DeepRoute-Driver 2.0, which uses 5 solid-state lidar and 8 cameras, can benchmark Tesla FSD, and even has better on-the-road performance, which costs less than $10,000.
On November 25, 2021, the Beijing Intelligent Connected Vehicle Policy Pilot Zone officially released the "Implementation Rules for the Pilot Management of Autonomous Driving Service Commercialization in the Beijing Intelligent Connected Vehicle Policy Pilot Zone (Trial)", and issued the first batch of domestic autonomous vehicle charging notices to some enterprises. Beijing has become the first city in China to explicitly recognize the commercialization pilot of autonomous driving "Robotaxi", and it also marks that the domestic autonomous driving track has finally ushered in the "second half" - the commercial operation stage.
Questions about how Robotaxi will change mobility and whether it can change the automotive industry will also be revealed. Robotaxi is not only aimed at replacing drivers, but also to completely improve the operational efficiency of the macro transportation system.
On December 29, Geely Holding Group announced that it has reached a cooperation with Waymo, an American driverless technology company, and its electric brand, Extreme Kr, will provide exclusive vehicles for the Waymo One driverless fleet and put them into commercial operation in the United States. The model is based on the KR intelligent mobility platform SEA-M (SEA-M) architecture, designed and developed by the Krypton European Innovation Center (CEVT).
This is just over half a year since March 2021, when Geely announced the official establishment of its new electric brand "Extreme Kr". As the world's leading self-driving technology company, Waymo is the most representative in the Robotaxi space, receiving $2.5 billion in a new round of funding in June 2021, with a latest valuation of more than $30 billion.
The Waymo+ Extreme Krypton cooperation model may have a profound impact on the future of the automotive market, with some OEMs reaping the opportunities for rapid growth from Robotaxi and others facing long-term bearishes.
Extreme Kr X Waymo
According to iCET's "2018 China Passenger Car Actual Road Driving and Fuel Consumption Analysis Annual Report":
China's passenger car annual mileage of 8,000 kilometers, an average of only 22 kilometers per day, from the official industry data can be seen, in the same year, the national automobile production and sales of 27.809 million and 28.081 million units respectively, passenger car ownership reached 240 million.
Considering that the above data includes the mileage of operating vehicles, the average annual mileage of non-operating vehicles will be further reduced, corresponding to 20 kilometers per day or less, or can be converted into less than half an hour of driving time, which can also be understood as the use efficiency of vehicles is only 2%.
From the perspective of the end of automatic driving, if the inefficient passenger car ownership can be greatly reduced, then a large number of traffic problems such as road congestion and parking can be effectively alleviated, RoboTaxi, RoboBus combined with the future high-specific energy battery pack, supercharge and other new energy technologies, will profoundly affect the choice of travel mode of all levels of society.
Assuming that the average daily usage efficiency of each RoboTaxi can reach 30%, the market demand for 14 passenger cars may be reduced by rough calculations compared with the 2% utilization efficiency of ordinary passenger cars.
But such an endgame may only benefit the head players of full-stack autonomous driving, and will also subvert the competitive landscape of OEMs. Not only traditional OEMs and new car-making forces will face the same problem, and the existing production capacity will exceed the future market demand in the long run.
Of course, RoboTaxi also needs a hundred flowers, if Ferrari also has Robotaxi, perhaps no one will refuse, even if it is 100 yuan per kilometer, after all, most people will not spend millions to buy it, nor will they spend thousands of dollars a day to rent it.
Therefore, when it comes to future car building, in fact, it may not really be a car, and new forces such as millet will also face complex choice questions, which may be single choice or multiple choice, and the title is not indicated.
GM is multi-choice, from the Super Cruise in L2 to ultra Cruise in L3 to the Cruise in L4, and its business model also deserves long-term attention.
Tesla is a single choice, full stack full self-research, firmly take the road of upgrading, but the proportion of end consumers who are willing to pay for private cars for automatic driving is not very high, and its financial report shows that in the Chinese market, less than 2% of car owners have chosen FSD, and how to increase this proportion, there is no effective measure at present.
How high is the threshold for high-level autonomous driving
1000TOPS hash rate may just be the beginning
If you want to define the full-stack capability of automatic driving, at least at this stage, the core should be computing power, but as automatic driving evolves to a higher level, the core will no longer be a simple computing power.
The 2022 hashrate ranking of advanced assisted driving vehicles T is based on public information
The demand for computing power in autonomous driving is endless,
In 2014, the original Tesla Model S used Mobileye's EyeQ generation chip, and the computing power was only 0.256TOPS;
By 2019, Model 3 began to use self-developed chips, and the computing power was increased to 144TOPS;
Then in 2020, WEILAI's newly launched ET7 uses 4 NVIDIA Orin chips, and the computing power exceeds 1000TOPS.
Players in the domestic autonomous driving chip track have made extraordinary moves in 2021, especially the horizon robot that is thriving in terms of financing amount and financing frequency, and its journey 5 chip released in the middle of the year has reached 128TOPS; Black Sesame Intelligence has also released a large computing chip Huashan No. 2 A1000Pro, with a computing power of 106TOPS; in addition, Xin Chi will launch a V9P/U with a hash rate of up to 200TOPS in 2022.
In terms of OEMs
First, Weilai Automobile announced the plan of self-developed chips; Zero Run Automobile directly released the vehicle-grade AI intelligent driving chip "Lingxin 01 CPU Processor" using Ali's PingtouGe Semiconductor Company's "Gentetsu C860"; Huawei, positioned at Tier0.5, has so far released four different computing power levels of MDC210, MDC300, MDC610 and MDC810, of which the strongest MDC810 hash rate can reach 400TOPS.
However, high power consumption is still a problem that needs to be continuously solved by high-computing chip platforms. Even if NVIDIA performs well on the track, the power consumption of its chips and computing platforms is still a big challenge. Based on the L5 level automatic driving computing platform developed by the main product Orin, although the computing power can reach an exaggerated 2000TOPS, the power consumption of 750W can also be borne by fuel vehicles, and for electric vehicles, it is equivalent to opening another set of air conditioners, which poses a considerable challenge to the current power battery capacity.
While domestic manufacturers are catching up quickly, they have made breakthroughs in vehicle regulation certification and mass production of L2 and L3 fronts. However, looking at the field of higher-level automatic driving, domestic head players are still in a large-scale lack of state.
In a series of interviews, a number of high-level self-driving startups have taken an evasive attitude towards how to evaluate the advantages and disadvantages of autonomous driving chips and selection criteria, but in the actual choice, the answer is very honest.
Combing the domestic and foreign Robotaxi and high-grade automatic driving tracks can be seen that in the global head camp, in addition to Cruise, Tesla is still self-developed, the rest of the preferred NVIDIA chip solutions, specifically including but not limited to: Waymo, Aurora, Argo, Motional, Baidu Apollo, AutoX, Momenta, Pony.ai, WeRide It's pretty much the same in the Field of RoboTruck.
In fact, for high-level automatic driving, computing power is not the primary consideration criterion, the important reason for NVIDIA's victory at this stage is that only Nvidia can provide a unified architecture and a unified software development environment for desktop, cloud and car terminals among chip companies in the current market.
In terms of car companies
Wei Xiaoli, the head of the domestic car-making force, has announced the use of NVIDIA Orin chip L3 models, the difference is the use of different numbers of chips. In the international market, Mercedes-Benz, Volvo, Hyundai and Audi have all announced that they will adopt NVIDIA's solutions.
With the continuous entry of new forces and new forces, the cycle of new models from design to off-line continues to shorten, and the automatic driving system is changing the traditional concept of car manufacturing from the perspective of development efficiency, so the integrated architecture and development environment have become the most productive selling point.
At GTC2021 on April 12, Huang Renxun released a new autopilot SoC Atlan chip, and the computing power of a single SoC can reach 1000TOPS, which is nearly 4 times higher than the computing power of its previous generation flagship Orin chip, and higher than the vehicle computing power of most current L4 level autonomous driving solutions.
In addition, NVIDIA also launched a complete autonomous driving solution "Drive AGX Robotaxi", which is not only a computer architecture and sensor platform for the automatic driving system, but also a full-stack autonomous driving platform that integrates automatic driving function modules and interfaces.
Nvidia is accelerating the expansion of technology originally developed for computer games and graphics into the automotive industry.
Omniverse has become an OEM digital twin
and the basic platform for simulation of autonomous driving companies
At NVIDIA GTC in November, the upgraded Omniverse platform was able to simulate warehouses, factories, physical and biological systems, robotics, and self-driving cars. The newly released Omniverse Replicator is a synthetic data generation engine for users to train deep neural networks, with two core applications:
Isaac Sim for general-purpose robots, DRIVE Sim for autonomous driving;
Drive Sim simulates a visual image of a "surround camera" mounted on an autonomous vehicle.
The virtual environments made up of these platforms and applications can be used to allow autonomous driving algorithms to train their own strategies.
The arms race in the chip field continues to escalate, and in October 2021, Qualcomm and Magna bid to acquire Tier 1 Vininger, and finally reached an acquisition agreement with Veninger. Qualcomm and its partners jointly acquired Veninger for $4.5 billion. In 2022, Qualcomm's Snapdragon Ride autonomous driving platform will be mass-produced and loaded, and will integrate Weininger's autonomous driving vision software.
For domestic autonomous driving chip companies, as well as autonomous driving technology research and development enterprises, NVIDIA's products and solutions must be learned even if they are difficult to benchmark, and how to obtain more comprehensive and rapid development capabilities through efficient self-research, cross-field cooperation, investment and mergers and acquisitions.
Variables still exist at the perceptual level
Lidar is still coveting the metacosm
As long as Tesla does not go to lidar for a day, the topic of the perceptual level of the endgame is still valid, and reading the lidar field in depth can also vaguely feel that although autopilot is still a long way from the ideal endgame, perhaps a rapid breakthrough in the field of segmentation can accelerate the progress of the entire process.
The perception layer solution that will dominate the mainstream market in the future will most likely include all-solid-state lidar and 4D imaging millimeter-wave radar.
According to the forecast data reported by Sullivan, the global market size of lidar in 2025 is 13.54 billion US dollars, which can achieve an average annual compound growth rate of 64.63% compared with 2019.
Referring to the relatively mature millimeter wave radar market, penetration is also gradually climbing.
According to Yole Développement data, the global vehicle-mounted millimeter wave radar market size is expected to grow from $5.5 billion in 2019 to $10.5 billion in 2025, with a compound annual growth rate of 11%.
The perception layer is critical for cars and autonomous driving, but the market isn't as big as it seems. In contrast, in the field of consumer electronics, Apple's single-brand Airpods will be a single category of humble Bluetooth headsets, quickly achieving a market size of $20 billion.
Considering the massive demand for 3D space modeling in the future metacosmology and spatial intelligence, perhaps in the future, every mobile terminal similar to a mobile phone needs a more powerful micro-lidar, iPhone12 may be regarded as an early metacostem device, and lidar This type of perception layer device will also see a real incremental market.
Lidar could be ubiquitous in the future
4D imaging millimeter wave radar currently has three mainstream technologies around the world that enable 4D point cloud capabilities, including:
1) Using existing Infineon, TI, NXP chips, through unique software algorithms and antenna design, high-power virtual MIMO is made on the basis of MIMO, so as to achieve a virtual number of antennas ten times on the basis of the original physical antenna number, and then greatly reduce the angular resolution;
2) The above functions are achieved by concentrating multiple-transmitter and multi-receptive antennas in one chipset;
3) Solve the above problems by using metamaterials.
Lidar is much more complex, both the latest technological field of autonomous driving at the perceptual level, and the area where capital has been betting on it in recent years and the market is relatively unclear.
Entrepreneurs in this field have said that they never know how competitors will solve problems. Sensors such as cameras and millimeter-wave radar, which have also existed and developed for many years at the perceptual level, are eventually installed in front of the car in a pure chip manner, of course, this is also a path that is more in line with the car regulations, so lidar will use the same route.
Vehicle-mounted lidar has inherent advantages and better performance in terms of ranging accuracy and anti-jamming compared to cameras and millimeter-wave radar. As low-level assisted driving upgrades to higher-level autonomous driving, and the autonomous driving system completely takes over the control of the vehicle, the role of lidar will also move from assist to dominance, and as the price continues to decline, the number of front-loading will also increase.
But to consider lidar, you have to go through quite a few dimensions. The different combinations of the two dimensions of the beam scanner and the detection system can yield a large number of feasible solutions.
Beam scanners can be divided into mechanical rotation, MEMS (micro-galvanometer), macro movement, Flash, optical phased array (OPA);
The detection system can be divided into incoherent measurement (TOF) and coherent measurement (FMCW);
The final product form can be divided into three categories: mechanical lidar, hybrid solid-state lidar and solid-state lidar.
Tier1 Bosch, the top 2 in the automotive field, has invested in tetravue in the United States in the field of lidar, as well as domestic Hesai Technology and Feixin Technology.
In a previous interview with Gesch Auto, Dr. Jiang Hongquan, venture partner of Bosch Group in Germany, said that lidar has the most technical space in hardware. Based on this, Bosch has successively invested in Feixin Electronics and Hesai Technology in the Chinese market. Among them, Hesai Technology focuses on the research and development of optoelectronic equipment, lidar and its core chips, while Feixin Electronics focuses on the research and development and manufacture of lidar for robots and unmanned vehicles. As a key technology in the development of automatic driving, even if L4 or L5 level automatic driving has not yet landed, lidar still has its market space, therefore, investment in Hesai technology is an inflection point to find the market, and investment in Flying Core Electronics is to find the inflection point of science and technology.
Now, looking at the entire technical field related to autonomous driving, and even the underlying technology, the gap between China and the world can be calculated in many tracks in "years", or even flat. What is even more gratifying is that many startups on the same track have formed a more rational competitive relationship by developing different technical paths.
Tao Zhe, the founder of Nasen Technology, who is breaking the monopoly of foreign Tier1 in the field of online control chassis, introduced: There is a new proposal in the industry at present, and the suppliers in the future supply chain are also called partners. In the era of fuel vehicles, the entire supply chain is pyramidal, one level suppressed one level. In the future, the automotive supply chain will be a network ecological structure, and the upstream and downstream will no longer be squeezed and mutually gamed, but a large ecosystem of partners will be established to achieve strategic synergy and joint development.
Possible Blue Ocean - RoboTractor
There are few more than trillion markets left
Putting aside the controversial topics about robots and Robotaxi, RoboTaxi, as the integrator of autonomous driving technology, realizes dimensionality upgrading through data accumulation, deep learning, and algorithm iteration, and can also distill models to reduce dimensions to multiple scenarios and low levels. Therefore, the Robo-family is also expanding, roboBus, RoboTruck, whether there is still a blue ocean, in the future will also be an interesting topic.
Agriculture is currently the field with the lowest degree of digitization, agriculture also represents the most basic just needs, and the improvement of its efficiency is of great significance to the national economy and people's livelihood and carbon neutrality. From a technical point of view, Robo's agricultural machinery will be able to achieve 24*7 work like a black light factory, which can maximize cost reduction and efficiency.
In addition, in the open agricultural scene, the organic supporting of photovoltaic, wind power, energy storage, and power exchange may be the first to open up the complete closed loop of new energy from output to consumption.
According to industry data, in North America, the market size of the agricultural automation field exceeds 5 trillion US dollars, and in recent years, the design thinking of agricultural AMR (Autonomous Mobile Robot) is changing rapidly.
More cross-border integration, such as deconstructing and integrating autonomous driving, tractors, and robotics to improve the productivity of the primary industry, has begun to achieve scenario-based landing, and the biggest variables in the future will also come from Robo+Tractor assembly like Lego.
Founded in 2018 in Seattle, Carbon Robotics, a weeding robot developed by integrating high-precision positioning, unmanned driving, robotics, and AI, weighing 4.3 tons, resembles RoboTruck, using NVIDIA's autopilot chip, using high-resolution cameras and AI recognition technology, will use 150W lasers in the farmland to accurately kill, and the error radius of its high-precision laser beam is only 3 mm. This self-driving weeding robot, which can work 24 hours a day, can burn 100,000 weeds every hour.
In contrast, domestic agricultural drones such as DJI and Jifei, whose main function is to spray herbicides and pesticides at ultra-low altitudes, have formed a mature business model, but the high residue and toxicity of herbicides and pesticides themselves are shortcomings that cannot be eradicated, and will be replaced by more environmentally friendly methods in the long run.
At the just-concluded CES, North American agricultural machinery giant John Deere showed off its first self-driving tractor, the 8R. Can operate 24/7, refueling every 8-10 hours. With 12 stereo cameras for 360-degree obstacle detection and distance calculation, the self-driving tractor ensures it is operating in the expected position by constantly checking its position relative to the geofence, with an accuracy of less than 1 inch, allowing precise sowing and harvesting of seeds without touching the steering wheel.
John Deere's driverless tractor 8R
According to its publicly available sources, the 8R automated tractor has collected more than 50 million images in field tests over the past three years. Each model is trained on hundreds of thousands of images, and its neural network can classify each pixel in about 100 milliseconds.
According to John Deere, the average age of the global agricultural population is already over 55 years old, working more than 12 hours a day. In China, especially in the vast rural areas, the aging problem has exceeded the global average, and it is difficult to reverse in the short and medium term in the future, from the perspective of labor structure, agricultural automation will inevitably become a trend.
At present, more than 80% of the world's agricultural and planting activities are completed by labor, and the average annual growth rate of labor costs in the agricultural field exceeds that of other fields such as industry and service industries, reaching about 10%, and the robot production line that has changed the industrial production structure, and the concept of black light factory, may be able to release greater productivity in the agricultural field.
From the perspective of infrastructure, although China's agricultural automation field started low, but the future will have a greater increment, according to China's agricultural census data, as of 2020, the number of tractors in the country reached nearly 27 million units, although far lower than the number of passenger cars, but has approached the number of commercial vehicles, and has a high frequency of use, short elimination cycle, a variety of subdivisions.
In the field of agricultural machinery in China, in addition to the lack of large-scale industry leaders in addition to "one tor", and there is still a relatively large gap between the level of agricultural machinery and the head camp of the international scope, automatic driving and robotics will likely help the agricultural machinery industry to achieve curve overtaking, or lane changing overtaking.
From a more macroscopic carbon neutrality perspective, RoboTractor can be associated with environmental protection energy technologies such as wind energy, solar energy, energy storage, power exchange, and new energy tractors; as well as artificial intelligence technologies such as automatic driving, robot automation, sound and image recognition, which can greatly reduce the energy consumption level in the agricultural field, form a complete environmental protection ecological closed loop, and will have a positive role in promoting the dual carbon process.
In the future, Ali's agricultural brain, Tencent's AI cucumber, and JD.com's smart farms may converge into a main line, or a new RoboTractor track may appear, and a super-large-scale market may be born.
Epilogue: The extension of autonomous driving
For innovative enterprises, autonomous driving provides opportunities to enter the market from new nodes in the industrial chain.
In the Ali metaverse defined by the XR Laboratory of Dharma Academy, an automatic apple picking robot appeared in the case of the highest level L4 (virtual-real linkage).
WEILAI crossed over and joined hands with NOLO and Nreal to create an exclusive VR headset and AR glasses for the car.
The autonomous driving track is super wide, and it is still necessary for domestic innovative enterprises to continue to explore more possibilities.
After reading it, click "Watching"