To teach a robot to avoid the tree and work in your yard while cut the grass, for example, is not as simple as teaching your dog to sit, or at least is not so simple with technologies used today in the robotics community.
Everything should be made as simple as possible, this is the idea. Replacing programming with training skills is more like teaching someone or somewhat doing something than writing code and having programming skills. But this is good as long as the operating system is generous enough to ensure the functionality on almost any system and for any robot.
An innovative operating system is about to change the way we learn robots doing tasks. The new way of learning is much simple than we expect, and the BrainOS is the software that replaces programming lines with trainings. After few repetitions when a human perform the task, the robot is able to start doing the task autonomously.
People are loving robots and many of these are active makers and hackers in the field. With this love in mind, I will explore all the BrainOS features as well as how the operating system can be used in robotics.
Inside, the BrainOS is not like any other operating system, but even so, the operating system has the ability to manage the sensors or actuators of the robot, as well as provides common sensorimotor processing such as manipulation or navigation for robot behavior.
The software was developed by Dr. Eugene M. Izhikevich, co-founder at Brain Corporation, and his aim is to produce low-costs service robots engineered for simple tasks and requiring significant expertise in programming.
Prototyping Brained Robots with A Healthy Dose of Focus
The BrainOS is designed to allow robots perform specialized tasks by building an environment that can grow with through training to perform tasks increasingly complicated. The training can hold days or weeks; all depends by the complexity of the job.
After Training: What’s Next?
According to software engineers that works at this project, a robot could be trained to use its manipulator and to navigate autonomously by detecting objects and know its position. In a real-world application, the robot should be able to use its manipulator for cleaning or sorting, and this should be an easily task. More than that, after intensive session of training, the robot may also be capable of taking out the garbage or cleaning up after pets.
Unique abilities cultivated by a trainer do not worth very much as long as the information cannot be shared. In a world of sharing everything with anyone, we should not be surprised if the software can be copy to new robots with the same design. With one click, a company could have an army of trained robots.
Dealing with BrainOS
Inside human brain works together about 86 billion neurons that processes and transmits information through electrical and chemical signals. Inspired by human brain, the BrainOS uses different artificial intelligence techniques based on artificial neural networks to provide a flexible artificial memory able to make connections and store information.
How do you train a robot?
The training session in the beta version of BrainOS can be made with a remote control, while an improved version will allow users to teach a robot like training an animal by observing the robot and reward the desire behavior.
For a social robot could be delicate to apply the training process due to a high complexity of tasks and environment. But, on the other side, many industrial companies are working hardly to add a learning type or another to their robots.
The bStem kit is the only one specialized hardware designed to run the BrainOS. The single board computer has an on-board Qualcomm Snapdragon S4-Pro processor and is featured enough to interface sensors, motors, WiFi/BT, stereo camera, and more.
Designed for hobbyists and students, the bStem kit comes with everything is needed to start building from scratch a robot with a brain.
In a demonstration, the researchers use a robot called EyeRover to learn how to navigate around a chair and other obstacles. The robot was able to replicate the maneuvers after the training session and yes, it is excited to read about robots able to have a digital replication of human brain, and definitely become electrified to teach a robot to do a certain task instead writing programming lines.