Almost 4 months have passed since Xiaomi officially announced the production of a car on March 30. In the past four months, except for news about Lei Jun’s visits to various car companies from time to time, there has not been much substantive news about Xiaomi Auto.
However, for a company that wants to seriously build cars, the preparation period of several months is actually not long. For example, NIO took a full three years from its establishment to the official launch of its first car, the ES8, not counting the preparation period before the company was established. Although Xiaomi has officially announced that it will build a car, it can only be considered in the preparatory period at best. However, everyone's attention to Xiaomi's car-making is obviously higher than that of previous new car-making forces.
No one knows what Xiaomi Auto will look like now, but a piece of news some time ago has slowly revealed the autonomous driving of Xiaomi Auto. According to media reports, Xiaomi recently acquired the autonomous driving company DeepMotion. Xiaomi’s autonomous driving may adopt a vision-based solution, configure high-precision maps, and support crowdsourced collection and update technology paths.
DeepMotion, also known as DeepMotion Technology, was founded in 2017. It mainly uses artificial intelligence technology to provide safe, reliable, mass-produced autonomous driving solutions based on vision. At present, technical solutions such as multi-sensor fusion perception modules, high-precision map automatic composition, and high-precision positioning have been developed.
According to information on the official website, DeepMotion’s four co-founders, including the CEO, CTO, chief scientist, and R&D director, all once worked at Microsoft Research.
Microsoft Research Asia is known as the Whampoa Military Academy for Chinese AI algorithm talents. Cao Xudong, CEO of Momenta (an autonomous driving company with a valuation of over 3 billion US dollars), and Ren Shaoqing, the former R&D director, both came from Microsoft Research Asia. , Ren Shaoqing later joined Weilai and served as Weilai’s VP of autonomous driving.
DeepMotion’s core team at Microsoft Asia Research focuses on deep learning and stereo vision research. By 2018, DeepMotion has matured the ability to produce high-precision maps of high-speed/express roads on a large scale, and the automation level of high-precision maps of more complex central urban roads has reached 90%.
According to DeepMotion’s official website, the company currently has four core products, including autonomous valet parking, intelligent sensing, high-precision positioning and high-precision maps.
Autonomous valet parking uses a surround-view fisheye camera, millimeter wave radar, ultrasonic radar, and IMU/GPS positioning module to realize that the driver is not in the car and the vehicle relies on the memory parking space for automatic driving and parking. Enter the corresponding parking space. Applicable basement types include multi-story basements, roadside parking lots, and open-air parking lots.
DeepMotion’s visual perception solutions include forward-looking, surround-looking and circumferential viewing, as well as multi-lidar perception solutions and laser/visual/millimeter-wave radar fusion perception solutions. Currently, it supports the sensing needs of different environments such as highways, urban areas, ports, basements, parks, and airports. It can identify vehicle drivable areas, traffic light detection, light pole detection, lane line detection, road traffic sign detection and recognition, and traffic sign detection. and recognition of static objects, as well as dynamic objects such as cars, trucks, buses, pedestrian detection, tricycle, motorcycle and other special-shaped object detection, and drivable area detection.
The high-precision positioning solution can currently achieve a stable real-time positioning accuracy of 10 centimeters in weak GPS signal scenarios such as urban areas, tunnels, and basements. At the same time, DeepMotion has developed a complete high-precision map solution from data collection, automated composition, interactive editing, quality inspection, etc.
It can be said that at the algorithm level of autonomous driving, DeepMotion’s technical capabilities are still very strong. After this acquisition, more than 20 people from DeepMotion may join Xiaomi, and its key personnel will assume the role of leaders of various modules of autonomous driving.
In addition to the acquisition of DeepMotion, Xiaomi has previously participated in the financing of two autonomous driving-related companies, namely Zongmu Technology’s D3 round of financing and Hesai Technology’s D round of financing.
Zongmu Technology is an autonomous driving (AD) and advanced driving assistance system (ADAS) technology and product supplier. It was established in 2013 and has obtained fixed-point contracts for L4 mass production projects from many domestic OEMs. , and has established mass production cooperation relationships with many domestic mainstream OEMs such as FAW Hongqi and Changan Automobile.
Hesai Technology is a manufacturer of autonomous driving lidar. Lidar is an important piece of hardware for smart cars to upgrade to high-level autonomous driving. It is reported that the funds from this investment will be used to produce hybrid solid-state lidar and develop automotive-grade lidar chips.
Therefore, in terms of intelligent driving, Xiaomi cars should have automatic driving assistance functions such as ACC adaptive cruise, automatic lane change assistance, lane keeping, and automatic parking. Models equipped with lidar can achieve L3 or even L4 level autonomous driving functions. After all, it will be a few years before Xiaomi Auto is officially launched, by which time autonomous driving may have become a standard feature of smart cars.
In addition to autonomous driving, what other functions does Xiaomi Auto have? Before announcing the construction of a car, Xiaomi has been getting closer to cars in terms of patents. Currently, the number of car-related patents has reached 834, and there are 446 related to autonomous driving, mainly focusing on wireless communication technology, driverless driving, and vehicle control. , direct-connect communication, Internet of Vehicles, gesture control, etc.
For example, Qichacha App shows that on July 6, Beijing Xiaomi Mobile Software Co., Ltd. disclosed the patent for "Vehicles and Control Methods, Devices, Terminals and Storage Media", with the public number CN113069774A. The patent abstract shows that the present disclosure belongs to the field of remote control technology, in which the terminal generates temperature adjustment instructions according to the current environmental information, and adjusts the hardness of the tire by adjusting the temperature of the tire, so that the hardness of the tire can adapt to different external environments and reduce the risk of tire damage. The wear and tear during driving enhances the driving performance of the vehicle.
When the temperature is high, the tires are softer, providing better driving comfort and grip; when the temperature is low, the tires are relatively hard and more wear-resistant. Adjusting the hardness based on tire temperature seems like a useful little feature.
With the support of Xiao Ai, the voice control system of Xiaomi cars should be one of the best in usability among current smart cars. By connecting to Mijia, you can also control smart home functions in the car.
Referring to the practicality and functional completeness of MIUI, the car version of MIUI may be closer to a smart tablet or smartphone in terms of functional interface, menu level design, and operational smoothness.
As for the model and price, Xiaomi should fully respect Mi Fan’s suggestions based on Xiaomi’s consistent thinking. Lei Jun once launched a poll on Weibo, which was "What car do you want Xiaomi Auto's first car to be" and "What do you want Xiaomi Auto's first car to cost approximately". Regarding car models, the voting results show that most people choose sedans, followed by SUVs, and the difference between the two is not big.
In terms of price, the most chosen ones are those within 100,000, followed by 100,000-150,000, 150,000-200,000, 200,000-300,000, and more than 300,000. The number of people choosing decreases in order, indicating that the vast majority Most rice fans still hope that the price of Xiaomi cars will be less than 150,000, which is also consistent with Xiaomi's long-standing cost-effectiveness strategy.
If Xiaomi’s first car has complete assisted driving functions (fully autonomous driving optional), good battery life, smooth operation, rich functions, and blessings from Xiaoai classmates, and the price is less than 150,000, then Xiaomi car, will you buy it?