autoevolution
Car video reviews:
 

Tesla's Autopilot Is Years Ahead of Competition Thanks to One Bold Decision

“If one person says you're a horse, they are crazy. If three people say you're a horse, there's conspiracy afoot. If ten people say you're a horse, it's time to buy a saddle,” goes the quote.
Mercedes-Benz E-Class autonomous test car 1 photo
If Tesla were to guide its business based on this advice, it would have started using LIDAR for its Autopilot self-driving suite a long time ago. All the other companies working on developing autonomous technology - with absolutely no exception, at least to our knowledge - are employing the laser-based scanning device in their range of sensors, and there's a lot more of them than ten.

In fact, the LIDAR is actually the centerpiece of their systems. In a Twitter post from a month ago, Henrik Fisker even talked about having five of them on his company's upcoming electric vehicle, the EMotion. He said that was the bare minimum for Level 5 autonomy, something Elon Musk might have a few words to say about.

Tesla has famously gone down the no-LIDAR route, and defended its decision on quite numerous occasions. A recent study shows it might have been right all along as the laser-powered sensor could be branded as "overkill" as far as self-driving cars go.

But what's the deal with LIDAR after all? It stands for Light Detection and Ranging, and it's the most accurate way to scan the environment. It sends out laser pulses that hit the objects nearby and bounce back. Based on the time needed for this, a computer can calculate the distance traveled by the light, and thus the sensor's surroundings are mapped.

The LIDAR is said to have an accuracy of 1.5 centimeters, which is about as much as the thickness of your pinky. But the real question is, do autonomous cars need such finesse, or are they fine with a slightly more broad picture of the environment?

The research carried out by a group of scientists from the U.S., Austria, Switzerland, and Singapore suggests they don't. They have concluded that video cameras alone can offer an accuracy of up to 10 centimeters. That's over six times less accurate than LIDAR, but it's still better than what the human eye can do.

Ten centimeters (or 3.9 inches) is no more than the length of the index finger of an adult. When you drive, you don't take your car closer than a finger's length to the other objects around, do you? So why would the AI need the extra precision?

The only drawback so far is that the tests have only been carried out at slower speeds. However, the team posits that the accuracy should remain the same at higher speeds as well if the proper software is used.

What this means is that Tesla is at least a few years ahead of its competitors. Not only does it have a much more affordable self-driving system, but it's already on the road bagging billions and billions of road data miles. At one point, the company would have gained enough information to teach its AI everything there is to know about driving, and that's when Tesla's Autopilot will finally be worthy of its name.

 
 
 
 
 

Would you like AUTOEVOLUTION to send you notifications?

You will only receive our top stories