Recently, a large number of start-ups are on the verge of falling off and are ready to rely solely on computer vision and radar to get autopilot. These companies include AutoX and Comma.ai. In addition, big companies such as Tesla also want to get rid of the burden of laser radar, light autopilot.
Of course, Tesla chose to go that route because Elon Musk had a small abacus: Lidar is now more expensive and less productive, but Musk wanted to get ready-to-go vehicles fully automated with software upgrades in the future.
These companies are so confident because of the recent advances in machine learning applications, particularly convolutional neural networks, that address computer vision. Artificial intelligence, especially pattern matching and computer vision technologies are constantly breaking through.
In addition, for start-ups, this is a huge bet. Big companies like Google are obsessed with LIDAR research, so they're loosening other solutions that give startups a tremendous amount of room to grow.
Of course, these companies may eventually lose nothing, but once successful, they can earn pours, while venture capital firms love to invest in such companies.
Three major automatic driving technology who will win the fight?
For these three key technologies, the trend now is very clear:
1. As time goes on, the price of laser radar will gradually decline, the performance will gradually increase. Ultimately, the high-end version can get a few hundred dollars, while the low-end version is cheaper;
2. Computer vision technology will evolve and ultimately meet the requirements of reliability. At the same time, with its use of high-performance processors and power needs will continue to decline;
3. The price of the radar will drop to double digits (in U.S. dollars) and the software used to analyze the radar data will become clever and smarter.
In addition, some new technologies are not well-predictable for the future, such as long-range long-wave infrared laser radar, new radar or Lidar alternatives that can directly radiate photons into radio waves.
Of course, even if new technologies are predictable, future winners must be companies that can combine multiple technologies. The question now is how these technologies are more effective.
The existing LIDAR has a problem that it is somewhat lower resolution, detection distance is not far enough and the price is expensive. Of course, there are a lot of problems with computer vision. At the moment it is not reliable enough, it needs the help of external lighting, and it requires high computer skills (which means expensive). The radar problem is also lower resolution, or even worse than the laser radar.
Below, we will introduce several effective technology combination programs:
Option One: high-end laser radar, supplemented by computer vision
The 32- or 64-line Lidar, favored by most teams, is very reliable in detecting obstacles to the road and is almost impossible to miss as long as it is within the detection range. However, once the distance is a little further away, they are likely to fail to tell exactly what the obstacles are.
Cnc Service,Metal Fabrication Service,Metal Laser Cutting Service,Cnc Machining Parts
Shenzhen Bergek Technology Co.,LTD , https://www.szbergek.com