Tesla vs. Self-Driving Car:



Automation is everywhere. Computers are taking over jobs in industries all over the world: agriculture, data entry, banking, and finance are all affected by the rising trend of machine learning and artificial intelligence. Now, the same technology seeks to take over the fastest machines in the world: our cars.

With car manufacturers investing millions of dollars a year into self-driving cars, they have solidified themselves as some of the most anticipated pieces of technology in recent memory. Almost all of those manufacturers seem to agree LIDAR is the way to go, but Tesla has other ideas.

Now, the real question is - whose approach is going to bear fruit?


Waymo’s approach

Waymo is an Alphabet-owned company that is widely considered the most advanced company dealing with self-driving car technology. It started out in 2009 with several tests on public streets and reached a ‘100,000-mile’ milestone three years later, in 2012.

Just three years after the first milestone later, the company would hit its million-mile milestone and quickly solidify its place as the leading self-driving company in the world.

Waymo’s technology is powered by a combination of LiDAR, computer simulations, high-definition maps, and powerful cameras. The problem several engineers and entrepreneurs, including Elon Musk, lies with the first three approaches to collecting data.

LiDAR stands for Light Detection and Ranging. It’s a piece of technology fires laser beams in a 3D space, and when those lasers are bounced back, the car can gather data from its environment. Inside the car, a machine learning model works its magic and helps the car to determine information such as the distance of various objects from the car, their height, speed, and direction.

LIDAR systems have been around for a while, and are so advantageous that scientists mapped Notre Dame with the technology, and are confident they can rebuild it back to its former glory.



So, why are Musk and his supporters so against LIDAR?

The main problem has always been they are very expensive. Waymo addressed this issue by manufacturing its own sensors, cutting down the cost from $75,000 to $7,500. Most people won’t appreciate an extra $7,500 markup on their cars all the same. Secondly, the simulations that computers rely on aren’t perfect.

Building a perfect simulation is a substantial engineering challenge on its own, but add on the need to handle vast amounts of data and you have an even larger issue. And, lastly, it’s argued the maps used to teach the on-board computer what the streets look like aren’t always reliable.



The Tesla approach

Fredrick Ty, a science and technology expert working for an online for NSBroker as well as in automobile engineering domain has a point to make. He says that Tesla’s approach is a minimalist one. Rather than load up their cars with radars and sensors, they opted instead for two cameras at the front of the vehicle. These do the same job as LIDAR, and, according to its advocates, it’s more accurate.

The argument is that cameras powered by AI are going to become so advanced that systems like LIDAR are going to be obsolete. Sounds pretty ambitious and promising, but we’ll have to wait for a few years before that’s verified.

The main benefit Tesla is likely to enjoy from this approach is the fact that it is a lot cheaper than LIDAR. This presents far fewer challenges in placing the technology in consumer-facing devices, as a top-rated research points out.

Before you ask, no, Tesla doesn’t have self-driving cars. What we’re talking about here is level 4 autonomous technology, at least. This would be such that the car can perfectly navigate on its own, but has to be geofenced to certain locations. The experts might tell you that it’s a lot harder to map the countryside than a city, for example.



Who wins the autonomous driving car race?

Ultimately, whoever wins the autonomous driving race is going to do so thanks to a single all-encompassing metric: data.

When a self-driving car collects data from its surroundings, the data is sent to a network of code that’s responsible for various processing and transformation operations. This code, referred to as a Convoluted Neural Network (CNN), crunches the data and identifies shapes, colors, velocity and other important data that are relayed back to the car. Using this information, the car can make the correct contextual decision - brake, speed up, or slow down.

Waymo engineers have been gloating over the fact that their cars have driven over 10 million miles on the road and 7 billion virtual miles in computer simulations. They hold a significant advantage over Tesla because of this.

Musk argues that computer simulations don’t take care of the more subtle ways people react on the road and unexpected issues like a pedestrian jumping in front of the car while it’s moving at full speed. Keep in mind that the Tesla approach doesn’t take care of these scenarios any better, either.

Another advantage Waymo has over Tesla is how portable their data is. The LIDAR system can be safely swapped from one car to another, so that, in the event of a car crash, for instance, none of your precious driving data is lost.

What Tesla has going for it is the fact that it already has hundreds of thousands of cars on the road collecting data for its self-driving car project. If the Tesla CEO is to be believed, this number will hit a million-plus by the end of 2020. And, true to Musk’s words, the camera-based system found on Tesla’s are a lot more affordable than LIDAR technology.



Conclusion

So, who wins the self-driving car race? Our bets are on Waymo. They started off earlier than Tesla and have collected more data than the latter can hope to match any time soon. Then again, it’s not written in stone. Musk has a lot of proponents and has surprised us before.