Stanford Raced Its Self-Driving Audi TTS Against a Human Driver
Have you ever wondered if Einstein could work with IoT? Or can you imagine Galileo working for space ships? Or maybe William Higinbotham playing PUBG with you? That would be something out of this world, right? Something really insane. They could have come up with things maybe that we could not. Maybe even today’s technological advancements would get a tough time dealing with such extraordinary minds.
Here in this article, we are going to explore one such news. News that kind of brings together our best in world technology with best in world experience. Enough of beating around the bush, let’s go straight to the point. Stanford University is doing research on self-driving cars along with a champion amateur racer. This research set a self-driving Audi TTS on a course against a real racing human.
No, no, do not go there, as mentioned on TechCrunch, this is no fascinating stunt but a real-life test. Moreover, it’s a warranted activity due to the sort of research it is. The researchers did not go participate in formula one racing if that’s what you are guessing. They went there as researchers and had it researched separately. Though nobody got hold of the actual lap timings, as far as we know.
Why this sudden need for speed in Stanford?
You must be really curious to wonder why such an extreme test for a self-driving car. A chunk proportion of the miles driven by these cars were at ordinary speeds. And almost all sort of obstacles encountered were equally ordinary. But what if the ordinary self-driven car faces a not so ordinary obstacle? Let’s say something challenging its ordinary friction bounds? Is it ready for such a thing? And what is the role played by AI and machine learning?
The journal Science Robotics published an article, starting with an assumption. Only a physics model is not sufficient or accurate enough for real problems. These are basically computer models that simulate the car’s motion in terms of weight, speed, road surface, and other conditions. But the thing with theory is it is either too idealistic or it is simplified according to our need. As we know errors when calculated with errors give out more errors. Hence, for something beyond the usual, it’s not enough.
According to the researchers, we are currently not equipped with hardware that can simulate a detailed result. Like if we need a simulator to simplify each wheel to a point or line during a slide. It is very significant that which side of the tire is experiencing the most friction. But we summarize results into an input and output. Then we could just feed that data into a neural network.
How will that help? Well, let’s suppose the car is fed with basic turning simulations. Like when the car is moving in may be Speed P and needs to take a turn Q, it knows how to make the move. But the problem is when it does not go that way or is a bit of line. Then we can have that fed input to help.
Then why racers with self-driven cars?
But that’s also quite specific. Under automation, our goal is to make machines that outdo humans or do not require their input for a better achievement. That is where these racers came into consideration. The Stanford researchers needed to compare the car’s performance with a human driver who knows from experience how to control a car at its friction limits.
The research team picked Thunderhill Raceway Park in California for the race. Against the human racer, they sent Shelley, a modified self-driving 2009 Audi TTS.
The research specifically focussed on parts with a lot of turn varieties and compared the opponents. Shelley has an advantage of consistency throughout the laps. But the human driver showed their experience and changed their line according to the car. And that is something the researchers are currently working on.
Therefore, we are still distant from our goal. Our goal of disarming ourselves in front of machines. Until that time comes, people will keep on experimenting. It will be both fun and exciting to watch machines made by us outrunning us.Tags: self-driving car, Stanford