Which is more reliable, autonomous driving or perception technology?

Musk, who claimed that L5 autonomous driving would be achieved in 2021, recently "bowed" to reality.

Ever since Musk started working on autonomous driving, there has always been a lot of buzz. He was the first to push new energy vehicles to large-scale commercial realization, the first to send rockets into space to build satellite Internet in the form of a private company... Recently, Musk publicly admitted in an interview with the media: autonomous driving is more difficult than imagined.

Musk has been the subject of controversy and criticism for years for his high-profile statements on autonomous driving. Most of the criticism comes from safety advocates, politicians and others in the auto industry. They say Tesla's sloppy approach to autonomous driving technology puts people in danger. They argue that overly optimistic promises about what these systems can or will be able to do encourage people to abuse them and cause crashes.

But some people believe that Musk is more willing to take risks than most other companies developing autonomous driving technology. His willingness to put unproven technology on public roads will accelerate Tesla's progress.

From the chip level, autonomous driving can be roughly divided into three schools: Intel, Nvidia, and Tesla. From the perspective of autonomous driving perception, Intel and Nvidia dominate LiDAR, while Tesla adopts pure visual perception that relies solely on cameras for identification. Wang Yu, executive deputy secretary-general of the China Productivity Promotion Center Association, pointed out in an interview with a reporter from China Electronics News that in non-extreme weather conditions, cameras are more accurate than radars, but if you encounter severe weather conditions such as heavy rain, fog, and heavy snow, it is difficult to rely on a single sensor to cope with it.

The safety issues that may arise from advocating pure vision are one of the main reasons why Tesla's autonomous driving is controversial. Recently, Tesla, which advocates pure vision, has welcomed a potential competitor.

Recently, Qualcomm announced that it will bid for Swedish autonomous driving startup Veoneer at a high price of US$4.6 billion. Veoneer's main products include autonomous driving assistance software Arriver, which can add functions ranging from collision warnings to parking assistance, and can further collect data through devices such as cameras and radars to monitor the surrounding environment, analyze it and provide corresponding response behaviors. If Qualcomm successfully acquires Veoneer, it will accelerate its research progress in the field of smart cars. Entering the field of smart cars is Qualcomm's focus at this stage. It is understood that Qualcomm has already used Qualcomm chips on 23 models of 17 car companies, and the latest autonomous driving solutions next year will be integrated into 5nm autonomous driving chips.

From the information revealed so far, Qualcomm is most interested in Veoneer's autonomous driving software Arriver. Arriver includes other driving assistance systems such as autonomous driving perception and driving strategy, which can help accelerate the implementation of Qualcomm Snapdragon Ride autonomous driving solution, which can choose three autonomous driving levels: L1/L2, L2+ and L4.

It is worth noting that Veoneer, which Qualcomm intends to acquire, has been working with Silicon Valley lidar company Velodyne since 2017 to sell various lidar sensors to OEMs and develop automotive-grade solid-state lidars that can be used on a large scale in L3 and L4 autonomous vehicles. In the long run, Qualcomm and Veoneer will have a lot to do in lidar.

As major manufacturers have entered the market, LiDAR has become an important technology needed to keep automakers ahead in the competition. In the past two years, a series of accidents such as "Tesla brake failure" and "Tesla crashed into people" have occurred frequently, and there are more and more doubts in the industry. Musk claimed a year ago that Tesla would achieve L5 fully autonomous driving in 2021. After various accidents this year, Musk finally admitted: autonomous driving is more difficult than imagined. Today, Nvidia and Intel continue to increase their investment in LiDAR to accelerate the implementation of their respective autonomous driving solutions; Qualcomm's acquisition of Veoneer also sends a signal of joining the LiDAR autonomous driving camp.

As the rival camp becomes stronger and bigger, how far can Tesla's pure vision go? Although it is still uncertain who will win or lose, judging from the current situation of Tesla, if it wants to be the first to realize the large-scale application of L5 autonomous driving, the pressure is increasing.