The world’s most valuable car company-the owner of Tesla, Musk has sprayed the key technology of autonomous driving more than once- Lidar , The speech is intense:
“Fools use lidar.” “Whoever uses lidar now, who is doomed, is doomed to doom!” “Expensive sensors are unnecessary. It’s like a lot of expensive accessories.” “Like an appendix. One appendix is not good-then, how about a lot of appendixes? This is ridiculous. You will understand.” Although for a maverick business genius like Musk, spraying any technology is not worth fussing, but hating lidar so deeply is really confusing. After all, lidar is a very core technology for new energy car companies like Tesla, which is destined to achieve fully autonomous driving in the future. because, Lidar is the eyes of autopilot , Used to help the car recognize all surrounding cars, pedestrians, obstacles, etc. The reason why Musk dared to be so “professional” is because Tesla did not choose the route of lidar, but chose another technical route: the use of high-definition cameras + visual recognition algorithms. The main reason for abandoning lidar is that Tesla has a deep enough accumulation in visual recognition technology on the one hand, but the more critical point is: Lidar is too expensive . Tesla, which keeps cutting prices, is obviously unwilling to bear more costs. However, other new energy car companies in the world have obviously chosen more precise lidar routes. The new car-making forces in China are more radical on the road of lidar + autonomous driving. On April 14, Xiaopeng, one of the top three new domestic car manufacturers (sha), launched its new smart car-Xiaopeng P5, and announced P5 will be the world’s first mass-produced lidar smart car . You know, only 3 days later, on April 17, another domestic smart car brand ARCFOX had already announced in advance that it would launch the world’s first commercial model Alpha S equipped with lidar. . In other words, Xiaopeng Motors rushed for 3 days and won the title of the world’s first officially released car company equipped with lidar. Although there is no use for eggs. In this way, suddenly like a spring breeze, the lidar was turned on everywhere. The Chinese New Energy Corps officially opened the curtain of the lidar + autonomous driving era. Moreover, what is more worthy of everyone’s attention is that whether it is Xiaopeng or Jihu, The lidar technology used is all developed by a local Chinese company , Livox, a lidar start-up company under DJI, provides Lidar for Xiaopeng, while the supplier of Extreme Fox is Huawei, which has not had a good time to transform everywhere recently. Congratulations, lidar, one of the most core technologies in autonomous driving, has been mastered by many domestic companies at the same time, and mass production and commercialization have been realized. Why does autopilot need lidar? LiDAR (LightLaser Detection and Ranging) is a radar system that determines the distance between the measured object and the test point by emitting a laser beam to the measured object, receiving the reflected wave of the laser beam, and recording the time difference. Under the “field of view” of the lidar, the surrounding objects are as shown in the figure below, which means that although there is no color and detail, the outline is clearly visible. Compared with the visual recognition method adopted by Tesla, because lidar does not need to rely on the clarity of the camera’s image and the algorithm to identify objects, it is completely physical ranging, so the judgment of surrounding objects must be more accurate, especially when the line of sight is out of sight. In a good dark scene, lidar is far better than a camera. At this time, the camera is likely to cause the algorithm to recognize the error due to the unclear object, and eventually lead to a car accident. The same is lidar, there are still good and bad points. Generally speaking, “how many lines” (see clearly, not how much money) is used as the evaluation of the accuracy of the lidar. The so-called number of lines means that the lidar has multiple transmitters and receivers in the vertical direction. Through the rotation of the motor, it can be How many laser lines are scanned? Of course, the more lines, the more perfect the surface contour of the object, but the larger the amount of data that needs to be processed, the higher the hardware requirements. In the animation above, you can actually see the concept of the lidar scan line more clearly. The high cost of lidar is mainly due to the complexity of multi-line lidar technology. The lidar technology itself is not very tall, and the 2000 yuan sweeping robot is equipped with a single-line lidar. Even the domestic head brand of sweeping robots, Robo, is self-developed multi-line laser radar based on the laser radar technology of sweeping robots, and looks forward to entering the field of autonomous driving in the future. . Compared with the solution of high-definition camera and algorithm, Lidar is too expensive. Therefore, if the smart cars launched by car companies do not need to achieve such a high degree of automatic driving, such as only assisted driving, there is no need to carry lidar at all. At present, autonomous driving is divided into 6 levels, among which, starting from L3, smart cars are considered to be capable of realizing automatic driving when the driver completely “collapses” in the driver’s seat (such as falling asleep). And it is generally believed that From L3 The use of visual recognition technology alone is no longer sufficient to support the accuracy of the recognition of surrounding objects when the car is completely separated from the driver. Lidar technology must be adopted. Therefore, whether it is the Xiaopeng P5 or the Polar Fox Alpha S, this time they are going for true autonomous driving. The Xiaopeng P5 is equipped with a 144-line lidar, while the Polar Fox Alpha S is a 96-line car-gauge lidar, both of which are high-end products in the lidar. Tesla, who disdains the use of lidar, has actually been hovering at the L2.5 level of assisted driving. Recently, the US media exposed two emails that Tesla wrote to the California Department of Motor Vehicles (DMV) at the end of 2020. Tesla acknowledged in the e-mail that Tesla is not currently fully autonomous driving, but belongs to the L2 automatic driving assistance system, the current software is not suitable for fully driverless operation, and will not make major improvements this year. And the California Department of Motor Vehicles here releases a ranking list of autonomous driving technology based on measured data every year. In the just announced 2020 autonomous driving ranking list, Google’s autonomous driving company Waymo once again topped the list, and the two Chinese companies AutoX and Pony.AI entered the top five. Two other Chinese companies, WeRide and Didi, also managed to squeeze into the top ten, ranking sixth and seventh respectively. Name, and Tesla is not on the list at all (abandoning the game). Of course, lidar is not a panacea. After all, Lidar cannot obtain the real image information of the outside world, but only scans and perceives the outside world through “lines”. Therefore, on the one hand, the detection distance is limited, which is much smaller than the camera shooting distance; on the other hand, the frequency of laser scanning by laser mines is much lower than that of cameras (P5’s lidar refresh rate is only 10Hz), the response is even slower. Therefore, the current mature autopilot technology uses a variety of sensors to increase the reliability of the car’s recognition of the outside world. It also requires visual recognition and lidar. For example, Xiaopeng P5 has a total of 32 sensors, including 13 high-definition cameras, 5 millimeter-wave radars, and 12 ultrasonic sensors (the accuracy of millimeter-wave radar and ultrasonic sensors is much lower than that of lidar, but some lidars fail. The scene can assist in recognition) and 2 car-level lidars.And a driving self-driving car is actually Under the combined action of multiple sensors that can perceive different distances, angles, and precisions, a 360-degree surrounding environment discrimination without dead angles can be realized . So why can domestic smart cars sold for 100,000 200,000 be equipped with lidar? Is the cost affordable? This is because, Once a Chinese company intervenes, no matter how high-end technology is, it will give you the price of cabbage. . At present, the high-end lidar originally sold for tens of thousands of dollars, Huawei has achieved the price below 1,000 US dollars, and at the same time it claims to be reduced to 200 US dollars in the future, Livox’s lidar has also achieved the level of several thousand yuan, and the key is also its own. Core technology and patents. At present, there are tens of thousands of high-tech companies such as Huawei and Livox that are attacking various core technical fields in the field of smart electric vehicles. However, there is still a serious and blind worship of American smart car-related companies such as Tesla and Google in China. It is useless to spray domestic electric vehicle companies or technologies. I personally feel that in the future, which country can lead the new era of automobiles, and who will die is not necessarily so absolute.
You must log in to post a comment.