Current robots are helpful, efficient, obedient and powerful, but they are also a bit limited. Most times these machines can only do what we program them to do, and programming is no easy task. Brain Corporation is changing robotics by creating eyeRover, a robot that can learn through experience, much like living animals.
We were lucky enough to see this technology in action at Uplinq 2014, where the cutest little robot managed to wow the whole crowd. The eyeRover was able to learn in the spot. It was taught how to avoid certain obstacles and how to follow certain commands.
It also helps that it’s a pretty good-looking robot. The eyeRover balances itself on 2 wheels, much like a Segway. It is completely 3D-printed and really feels like some kind of pet/animal when in action.
Training instead of programming
What makes this technology interesting is that they are treating robots much like animals. These are not perfect and will not be allowed to drive a car for you, for example. This is because, much like us, they learn through experience and act like thinking beings.
The eyeRover runs on BrainOS, a system that can get distracted and make mistakes. It also learns more with experience. The more it does something, the more efficient the robot will be at it.
Why is this any good? Unlike the self-driving car industry, other industries will find such robots to be perfect for their applications. Not to mention, robot learning is much easier than programming.
In traditional programming, one would tell the robot every situation and command it to react a certain way. For example, telling it to evade a bucket would require us to tell the robot how many degrees to turn to the right, what distance to travel and so on – for every single action.
How does BrainOS learn?
Want to train your robot? You won’t need to spend hours coding before it can take a single lap around your offices. BrainOS can learn via supervised learning or reinforcement learning.
Supervised learning is when the user controls the robot with a wireless controller. This shows the robot what the desired behavior is. You show them what you want them to do, and they will then replicate the action.
Reinforced learning is much like what we know as classical conditioning. You simply observe the robot and “reward it” when it takes a desired action.
I can see this type of technology becoming huge in the future. Brain OS would make for great maids, for example. You can teach it to clean certain things when they do not look a certain way. Maybe you could teach it to pick up the dirty clothes off your kids’ bedrooms!
Those are only a couple examples of how Brain Technologies could shape the future of robotics. The possibilities are endless, and learning robots is really a dream come true. Unless they start thinking too much and overpower us… that wouldn’t be great.
Check out the video and learn all about BrainOS and the eyeRover! It’s one hell of a demo.