Modelowanie 3D Mechanical

AI bicycles that "think like humans"

Przez Rose Barfield 5 min 8 sierpnia 2019
AI bicycles that "think like humans"

Forget apes and robots taking over, the future dictatorships could be run by super-smart, self-driving, AI bicycles! A bicycle has been fitted with the latest in AI technology which allows it to learn "like a human". Suddenly Knight Rider doesn't seem so farfetched.

Bicycles that think like humans

A team in Tsinghua University, Beijing, China has combine technologies that allow a bicycle to stay upright....

Their incredible developments in AI, mean that this bicycle that can follow you around and listen to your commands. The new approach to AI combines two previously "incompatible" methods to create a system that can make decisions and learn in a way similar to the human brain.

Watch video on YouTube

By now we've all heard about self-driving cars, but a self-driving bicycle and one that can learn, boasting "true AI" no less! It all feels very science fiction!

Staying up

As any child learning to cycle will tell you, bicycles are highly unstable. This is thanks to their high center of gravity and only having 2 wheels. In addition, bicycles don't travel on perfectly flat ground, without obstacles, although it would be nice if they could! Unlike motor vehicles, bicycle architecture covers a much wider variety of locations; alleyways, mixed usage paths, bus lanes, country roads, etc. All of this makes designing a bicycle that can stand up on its own really tricky.

Watch video on YouTube

Several studies have attempted a similar thing using various techniques. Most notably: Murata Boy. Murata boy is a rather adorable, cycling robot able to cycle along a 2cm wide beam and even maintain balance at a complete standstill. He does this using a gyro sensor that sends signals to a rotating disc in his chest. He is also equipped with ultrasonic sensors and an onboard camera. However, this balance system is only successful due to Murata Boy's small size.

Remote control bicycles

A remote control, self-balancing bicycle was developed in 2018, by students at the Automation Experimental Teaching Center of Tsinghua University in Beijing. Probably the first major innovation in the history of the bicycle since the early 1800s. And no, unlike the Dutch, Google, self-driving bicycle, this time it's not an April Fools!

This bicycle uses a control momentum gyroscope, pretty similar to the one you would find on a multi-copter drone or even a spaceship. The downside is that a gyroscope has a high energy consumption. There is also a danger of the bike doing a wheelie! Although personally, I think that's kind of cool. Maybe the next step is a self-driving unicycle?

Tsinghua University isn't the only one at the self-driving bicycle game. There is also a driverless bicycle developed by Cornell University, although theirs looks a little more like sputnik.

AI Bicycle

The next step was to fit the bike with the latest AI technology. It can now follow you around and even obey simple voice commands, such as, "left" or "straight". It feels rather like a well-trained puppy.

The bicycle can simultaneously track and identify moving targets and avoid obstacles, cross speed bumps, maintain balance, follow voice commands and move in the correct direction. It needs to simultaneously compute balance, dynamic sensing, target detection, tracking, obstacle avoidance, understand speech, and make decisions. To do this requires a staggering 1,000 calculations a second!

Whilst you might be thinking "well what's the big deal a driverless car can already do that?", think about this: the speed and efficiency that the new AI can make these calculations is much faster than traditional methods. This is the really important part! The team also consider this "just the beginning" and have high hopes that they will be able to use this system to "learn on-the-fly".

Computers thinking like humans

Until now most AI has been very good at a single task but this requires thousands of simulations and the "intelligence" demonstrates almost no flexibility. For this reason, scientists have been searching for a way to make computers think and learn more like humans. This is known as AGI (Artificial General Intelligence) or "full AI" and is the basis of many sci-fi concepts. Man considers it the "holy grail" of AI.

In the pursuit of AGI scientists typically followed one of 2 "incompatible" paths: a neuroscience-based or a computer science-based approach. A neuroscience-based AI system seeks to mimic the human brain as closely as possible. In a computer science-based AI system the computer runs a learning algorithm.

Combining the systems

The two systems are very different. Communicating between the two is a bit like trying to translate between English and Chinese, it will never be perfect. They use a different language, different encoding, and even different storage methods.

Tsinghua University have achieved the unimaginable: they have successfully found a way of combining these techniques into one chip. It takes the form of a "heterogeneous fusion brain chip" known as the Tianjic chip (Meaning Ten Movement).

Will computers take over?

People pay attention to this issue because we are worried about whether or not humans will be destroyed as sci-fi movies say. In fact, what we can destroy humanity has already been created, that is, nuclear weapons, but why is it not destroying humanity now? Because we master it, we can control it. Like brain-like computing, strong artificial intelligence, and artificial general intelligence, we believe that humans can make good use of our wisdom to regulate its path of development so that it can benefit us and minimize those risks. -- Shi Luping, tianjic chip developer

Further reading

A self-balancing bicycle that uses steering to maintain balance

Another example using steering to maintain balance

The gyroscopic technology behind the self-balancing bicycle

10 marca 2024 6 min

The path to Scan to BIM automation with BricsCAD® BIM V24

The new automation tools included in BricsCAD® BIM V24 make the Scan-to-BIM process easier and more efficient! If you want fast and accurate reality capture of existing objects, these tools will be a great addition to your arsenal. We've added tools like the Point Cloud Classifier, Point Cloud Detect Rooms, Point Cloud Fit Rooms, and Normal Calculation for structured point clouds to help you create digital twins for quantity takeoff, create accurate building plans, and with facility management.

Śledź nas w mediach społecznościowych