Nowadays, the car is not like before, and it has been stable after mastering the "three major pieces".
If there is no layout of lithium ore, batteries, chips, autonomous driving, these upstream and downstream industries, the degree of difficulty in building cars will increase relative to competitors who have upstream and downstream industries. Xiaomi, which announced its car manufacturing in March last year, also began to pave the way for its own car-making, invested in the chip industry and the autonomous driving industry, in order to better have self-developed technology.
Among them, xiaomi's investment in longitudinal technology and the L4-level autonomous parking + smart city solution developed by Xiaomi seem to let us find highlights from the many technology companies invested by Xiaomi.
Autonomous parking, L4 level first puzzle?
There are a total of three technology products under the umbrella of Zhongmu Technology, namely intelligent driving system, new energy vehicle wireless charging system, and smart city solution. All three products are the future direction of autonomous driving, and all three products have automatic, guided and interconnected labels on them.
The so-called intelligent driving system of Zhongmu Technology, which includes the L4-level autonomous valet parking system, has been tested on test vehicles such as FAW Hongqi and Changan Automobile, and the functions can be seen in the released video that can already realize autonomous parking and autonomous driving. We can understand autonomous parking as the first piece of the puzzle of L4 automatic driving, and the two major plates of automatic driving are nothing more than "driving" and "parking".
The topic of autonomous parking has developed from APA to RPA to ADP, of which APA and AVP are the more mainstream functional products in the market. First of all, whether it is L3 or L4 level automatic driving, it requires a lot of computing and strong learning ability of artificial intelligence algorithms to cope with the changeable, strange and complex real environment, parking and driving are the same.
From the early stage of the development of AVP function, the sensing hardware used in supporting the use of ultrasonic radar, front-view camera, and surround view camera constitutes a fusion perception capability. The underlying logic is briefly discussed, the side APA ultrasonic radar detects the parking space, the algorithm relies on fusion perception to calculate the parking route, and the UPA ultrasonic radar determines the distance between the front and rear vehicles, and finally completes the autonomous parking.
Of course, this is only the ideal use case, in fact, the future development direction of AVP also needs to open up the three interconnection solutions of the car end, the field end and the cloud. The solution of Longitudinal Technology is more similar to the solution of the car end and the cloud, and there is no choice of high-cost field-side transformation solution.
The AVP Gen.2 vehicle-side solution used by Longitudinal Technology has 4 high-definition surround view cameras, 4 4D millimeter wave radars, and 12 long-range ultrasonic sensors. This set of equipment and algorithms allows vehicles to rely on 4D millimeter-wave radar to complete AVP functions in the event of visual perception failure. Among them, 4D millimeter wave radar is the highlight configuration of the perception device, with point cloud imaging function, which has a good performance in the accuracy of ranging and speed measurement, and is lower than the cost of lidar.
The car end has advanced perception equipment, which is the first premise of the car end solution, and the longitudinal view technology also provides crowdsourced maps, and after completing more maps, the car side can be applied to multiple scenes, and the combination of the two has higher requirements for the algorithm.
The underlying logic of the AVP solution in the cloud is that the vehicle park cloud uploads the data and information from the vehicle terminal sensor data and information synchronously to the cloud, and the cloud calculates the output instructions to control the vehicle. The advantage of this solution is that the performance requirements of the vehicle-side processor are reduced, the field-side transformation requirements are reduced, and the difficulties are in the three-party system communication logic and the reliability of communication.
The above two schemes are currently the most feasible solutions, low cost and conducive to promotion. The ultimate direction of AVP may be more inclined to the field side, which can also be understood as the V2X interconnection function in a certain area, but the field side is a high-cost choice.
The reason why the field-end solution is costly is that it needs to embed cameras, lidar, and high-precision maps in the park, which is equivalent to giving the car park a certain perception ability, so that the information calculated by the terminal is transmitted to the car end. The field-end parking lot jointly created by Bosch and Daimler is such a logic, the advantage is high safety and high precision, but it is not easy to promote the high cost of transformation.
Low-speed dynamic V2X, is it easier to achieve?
The three technology products of the above-mentioned Longmu Technology all have the function of "interconnection", the car end and the cloud can realize the interconnection, the car end and the wireless charging ground end device can be interconnected, and the higher-level development direction of the vertical view technology also gives a set of "smart city solutions".
According to the information given on the official website of Zhongmu Technology, this solution relies on the high-precision map in the cloud to have the functions of roadside parking management, charging management system, induction management system, reverse car search system and reservation parking. The above functions are more inclined to be implemented in static or low-speed scenarios, and the current development status of V2X is still mostly in the testing stage, because the regulations and technology are still relatively immature, mass production may take some time.
What kind of effect can the solution of Longitudinal Technology achieve in the end (logical analysis)?
Roadside parking, the realization of the vehicle end and infrastructure interconnection (V2I), showing the roadside vacant parking spaces on the map; charging management system, the above-mentioned wireless charging ground device, is also a kind of V2I interconnection; reverse car search system, to achieve the V2N interconnection between the vehicle end and the network (mobile device); induction management system, the above-mentioned depot cloud interconnection (V2N), for autonomous parking/driving induction.
Although the use scenario is concentrated on low speed/static, it has the interconnection function of V2I and V2N, and since there is no open environment and high-speed dynamic use scenario, the interconnection relationship and function of vehicle-to-vehicle (V2V) and vehicle-pedestrian (V2P) are temporarily not mentioned in the product introduction of Longitudinal Technology. I think that opening up all the functions of V2X is not a one-time thing, and it is the same step-by-step development model as automatic driving.
Supporting the above functions is a set of LBS enhancement service platform, LBS we are most accessible function is "real-time bus", providing bus route planning, real-time information and so on. If it is put on an autonomous vehicle, the service platform needs to complete the link 3D engine to ensure that the vehicle recognizes the sense of direction during navigation; the vehicle guide/step guide engine needs to provide the car or user with search functions, path planning and other functions.
The above, at present, is still a series of concepts and planning of V2X, and several difficulties are discussed.
First, V2X may bring about network security problems, take Tesla as an example, hackers can get a series of controls such as vehicle positioning and unlocking from the mobile phone app intrusion (can not start or drive); second, the technical support of V2I and V2N needs to be completed through the upgrading of supporting facilities, if the large-scale landing, then costs and standards are problems that need to be faced; third, if V2X equipment is added, it will inevitably directly increase vehicle costs.
Also doing 5G and V2X interconnection is Huawei, although Huawei does not directly build cars, but also can provide solutions for automatic driving systems, in 2019 Huawei has begun to work on 5G + V2X technology research and development.
summary
L4 level autonomous parking and V2X interconnection technology are the future development direction of smart cars, and Xiaomi has begun to lay out this industry from the beginning of car manufacturing. Xiaomi and Longitudinal Technology do not have the strength to do the above L4 level autonomous parking and V2X functions?
None of us can give a definitive conclusion until it really lands. However, compared with the dynamic V2V and V2P, static or low-speed V2I and V2N in a certain area are easier to achieve, not to mention that Xiaomi also has 5G mobile devices and 5G base stations. L4 level of autonomous parking, the biggest difficulty at present is the applicability of a larger range (multi-scenario) and the optimization of the algorithm, according to the results of the current test, the final realization effect has reached a high level.