Is a self-driving car safer than a driver's car?

Automakers and technology companies are racing to get the first self-driving car. This is really a "battlefield" in the future where the cars drive popularly and gradually replace people.

But is self-driving cars really safe, and can replace current drivers? Business Insider will help us answer this question.

Ford, GM, Tesla, Lyft, Google, . all plan to bring some self-driving cars ready for commercial use in the next 5 years. According to Goldman Sachs, it is estimated that unmanned cars can account for up to 60% of car sales in the US by 2030.

Companies all have their own reasons to invest in this technology, but all agree that one of the biggest benefits of self-driving cars is a higher level of safety.

At this point, you will definitely ask, how can a car without human control, no steering wheel, and no brake pedal be safer than the cars we have? The current?

Yes, the answer lies in the technologies used in these self-driving cars.

Own many sensors

Self-driving cars have a lot of sensors.

Self-driving cars are designed with the ability to recognize the world around "like superman" . They use dozens of sensors to collect environmental data so they can operate in a constantly changing environment.

Basically, self-driving car companies are really trying to simulate how people drive through the use of sensors.

"As a person you have senses, eyes, ears, and sometimes touch, you will feel the way. So these factors play an input and then will be put into the brain, the brain will decide how to control the legs and hands when needing brakes, pressing gas and driving, so on a self-driving car you have to replace these senses , senior director of Nvidia's automotive business, shared with Business Insider.

Some sensors used on self-driving cars include cameras, radars, lasers and ultrasonic sensors. GPS and map technology are also used to help the vehicle determine the exact location.

All of these have different strengths and weaknesses, but they basically allow a lot of data to be put into the car.

And a special brain

In fact, the Tespilot system Tesla uses is Nvidia's Drive PX2.

All collected data is then fed into the car's computer system, known as the "brain" , and processed so that the vehicle can make decisions.

One of the leading companies building brains for self-driving cars is Nvidia . In fact, the Tespilot system Tesla uses is Nvidia's Drive PX2 . This is the company's latest computer system for self-driving cars.

Drive PX2 is a powerful computing platform that uses a series of Nvidia chips and software aimed at bringing all data from sensors on self-driving cars to build a three-dimensional model for the transport environment. onion of the car.

"In the car's brain, it's almost like a video game. We need to recreate the world in a virtual 3D space," Danny Shapiro added. To do this, Nvidia and other companies develop unmanned cars using technology called maching learning.

So how do they learn?

To train a self-driving car, you will first drive hundreds of miles to collect data for the sensor.

Machine learning is a way of teaching algorithms with examples or experiences that companies currently use. For example, Netflix and Amazon both use it to offer referrals and advertisements based on what you have viewed or bought in the past.

So, to train a self-driving car, you will first drive hundreds of miles car to collect data for the sensor. You will then process that data in a data center that identifies the objects captured by each frame.

"Initially, computers didn't know anything. We have to teach it. If we want to teach cars to recognize pedestrians, we will take images of pedestrians, millions of images of riders. Because they look different, the more data is put in, the more cars can recognize, and we do the same with people riding bikes, cars, trucks, and we do that on "Different times of the day and at different weather conditions. Basically there is infinite ability to build a memory and understanding of what the car faces," Shapiro said.

In the future, with the continuous development of technology and artificial intelligence, we will have smart self-driving cars and almost no weaknesses.