Tesla, do you want to compromise with lidar?

2024-05-09 07:11 Source: Gesch Motors
Tesla has always adhered to a purely visual approach in the field of intelligent driving. Elon Musk, CEO of Tesla, has a negative attitude towards the laser radar sensor. He once mocked that "only fools can use laser radar".
However, recently Luminar Technologies, the US manufacturer of lidar sensors, revealed that Tesla was its largest lidar customer in the first quarter of 2023, with the purchase amount "accounting for more than 10% of the quarterly revenue". According to the financial report data of Luminar, the company's revenue in the first quarter was $21 million, which calculated that Tesla purchased more than $2 million in lidar products in the quarter.
This news is very surprising. It seems that Tesla's position on laser radar technology has changed. Is Tesla going to compromise with lidar?
   For self driving taxis?
Analysts from the Gesch Automotive Research Institute predicted that Tesla would spend $2 million to purchase lidar, probably for testing or other purposes, and would not give up its long held "pure vision route".
Musk once said frankly in 2019 that "anyone who relies on laser radar is doomed to fail, and expensive sensors are unnecessary". But in fact, with the continuous maturity of technology and process, the price of lidar has continued to decline in recent years, from more than 100000 dollars to hundreds of dollars.
Among them, Luminar is the first enterprise in the world to launch a $500 low-cost lidar scheme. Its new generation Iris+lidar was released last year, which is not only cheaper, but also thinner, and can be seamlessly integrated into mass production models. This product was delivered to major customers last year and is expected to be delivered to mass production models in 2025 and beyond.
Due to the reduction of cost, laser radar is rapidly popularized in the field of smart cars. In the Chinese market, the multimodal fusion perception technology of "vision+radar" has become the mainstream, and most car enterprises, such as Xiaopeng, Weilai and Interworld, have adopted this route. Tesla has become one of the few new energy vehicle enterprises that adhere to pure visual perception.
From the purchase amount of 2 million US dollars, Tesla can only purchase 4000 Luminar affordable lidars at most, which is far from meeting its product loading needs. After all, Tesla's current global monthly sales have exceeded 100000 vehicles.
In fact, except for cameras, Tesla does not carry even the basic millimeter wave radar and ultrasonic radar. Tesla executives have explained that "additional sensors will increase system complexity and cost". Musk advocates the concept of "fewer parts, better". Tesla believes that having a team that collects more data is more important than having more sensors. If the vision system can complete the driving task, there is no need for redundant sensors and related costs.
However, the latest news shows that Musk will release an autonomous taxi on August 8. Foreign media speculated that this might be related to Tesla's sudden emergence as Luminar's largest customer.
   Is vision more reliable than lidar?
Musk believes that the pure vision perception scheme is more reliable than the fusion of laser radar and other sensor schemes. He often questioned the radar fusion scheme on social media, such as "it is difficult for radar to see villains, but Tesla's vision is very obvious" "When the radar and vision are inconsistent, the vision has higher accuracy, so the pure vision ability exceeds the radar fusion scheme."
However, there are opinions in the industry that the reliability of purely visual route is questionable when dealing with complex road conditions and bad weather. For example, in foggy weather, the reliability of the vision system will be affected. However, the video of driving in fog shared by Tesla owners shows that the AP auxiliary driving function can work normally, and can detect the parking ahead and the oncoming vehicle.
But not everyone can achieve the pure visual route. The former senior director of Tesla AI once mentioned that "pure vision can accurately perceive depth, speed and acceleration information. It is difficult to achieve pure vision, and a lot of data is needed." Because of the high threshold and barriers, only a few domestic car companies, such as extremely few, follow Tesla's footsteps.
The latest news shows that Tesla FSD (fully automatic driving) is about to enter the Chinese market. FSD has evolved to V12 version, which can realize navigation assistance, automatic lane change, automatic parking, intelligent calling, signal recognition, urban road automatic steering and other functions, and has been promoted in the United States and Canada. At present, FSD sells for $8000 in the United States.
A few days ago, Musk made a flash visit to China and met with a number of senior Chinese officials to discuss issues related to the launch of FSD in China. Tesla APP shows that the official purchase page of FSD has changed from "later launch" to "upcoming launch". At the same time, Tesla has reached cooperation with Baidu Maps and passed the national automotive data security compliance requirements. It is speculated that Tesla FSD has entered the countdown to China.
According to the latest report of China Daily, Tesla also proposed to land "driverless taxis" in China. In this regard, China may first support domestic testing and demonstration, but has not yet fully approved the full implementation of FSD.
However, it can be predicted that once FSD enters China, it will certainly form a "catfish effect" in the automobile market and contribute to the development of domestic intelligent driving technology. However, it still needs time to verify whether it can promote the whole Chinese new energy automobile industry to take a big step forward, as its domestic models came into being in China that year. (Gaishi Xiangxiufang)

For more information or cooperation, please follow the official WeChat of China Economic Network (name: China Economic Network, id: sourcecn)

View the rest of the full text
(Editor in charge: Guo Yue)