Tesla has just announced that they’ll be switching all of their new Model Y and 3s ordered after May, and later their Model X and S’s, to use new technology for their autopilot called Tesla Vision. This replaces radar, which is what Tesla and most car companies have been using for their driving assistance and is based on cameras, rather than lasers creating maps which is what Lidar uses. While all car companies are transitioning away from the radar, everyone else is moving to lidar, except for Tesla and Elon, who famously said that Lidar is a “fool's errand.” But how do all of these work and why is Tesla so stubborn to use LiDAR?
Radar has been used in cars since the 1980s, and works by sending an EM wave from a transmitter which reflects any objects back into a receiver’s antenna and tells the processor how far away and where any obstacles are from the car. Nowadays, more and more cars are actually moving to use radar as a supplement to more advanced technology, like lidar and vision-based ones instead of being the primary technology. lidar works just like radar where it sends out beams of laser which bounce off any objects, but it also uses this data to make a 3D map. It was first used by the US military to track the satellite and lunar movement in the 1970s and 80s and it has gotten a lot smaller, but is still expensive and bulky, which is one of the main reasons Tesla has chosen to not add it to their cars. Speaking of Tesla, the third driving assistance technology is cameras, and Tesla’s new Vision is solely cameras, without radar or lidar. Camera-based technology like Vision works by sending the video data to the computer's AI, which can recognise and help avoid obstacles with its machine learning.
So why does Tesla not want to use LiDAR when it’s an established and reliable tech, compared to relying on cameras. It’s because of a couple of reasons like the fact that it has different advantages and disadvantages and takes up a lot of space. If you’ve seen pictures of Waymo cars, which do use LIDAR, you’ve probably seen that they aren’t the most attractive of cars. Seriously, they have sensors everywhere as if they’ve got chicken pox and that’s without mentioning the huge module for the sensors on the roof. In comparison, Tesla wants to market electric vehicles and self-driving as the future to ordinary people, and giving their cars sensors that bulge out and aren’t attractive isn’t exactly going to help their case. Plus, since LiDAR is 3D mapping its environment, it needs a lot of processing to store and keep up to date, while developing a database for the camera's AI is much easier. Also, LiDAR modules are very expensive, costing up to $10,000, and even when they are mass-produced like Waymo is doing, they still cost around $7500. In comparison, Tesla’s cameras cost in the hundreds, which will be very important since they will have self-driving in all their cars, including the cheap ones, so need to make it cheap. In contrast, other car companies can put LiDAR into their highest end most expensive cars so don’t need to worry about costs. Finally, Tesla believes that they can train their cameras to become just as good as LiDAR through lots of data and experience. When the cameras are trained more, they’ll be able to detect people and objects and even read signs better, which is a huge advantage over lidar. Now you might say that it’s better to use LiDAR and cameras together, but Tesla argues that the data from radar and LiDAR interferes with the cameras and that LiDAR will be dropped after a while anyway, so there’s no point in putting it in now.
In conclusion, there are a lot of reasons for Tesla to use vision-based technology over LiDAR and radar for the future of their self-driving cars. A good comment I saw was that roads are designed for human vision, not lasers, so when Tesla's vision gets more developed, it will have another advantage over LiDAR. I believe that Tesla is correct about cameras being the future instead of LiDAR, but since there haven’t been any tests on how good vision is since it’s just been implemented, we don’t know for sure if Tesla’s gamble has paid off. At the end of the day, these systems are designed so the people in the car are safer, so only time will tell if LiDAR will stick around or if Elon gets the last laugh.
If you want to learn more about Tesla Vision, LiDAR and Radar, you can visit these sources I used.
Sources:
https://www.ff.com/us/futuresight/what-is-lidar/
https://towardsdatascience.com/why-tesla-wont-use-lidar-57c325ae2ed5
https://medium.com/0xmachina/lidar-vs-camera-which-is-the-best-for-self-driving-cars-9335b684f8d
Comments
Post a Comment