Machine learning has been taken to a whole new level at the MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). The CSAIL and Boston University have developed mind-mapping techniques that can help to train robots for performing industry tasks better than ever before!
The factory robots that are custom-made to work on a particular assembly line task don’t often make mistakes, but when they do, it costs the company millions. Thus, training them to near perfection is essential, which would require a streamlined process to teach a robot, like Rethink Robotics’ Baxter, how to do its job properly without having to stop and reprogram it constantly.
Researchers have achieved this feat through their tech-savvy yet affordable and flexible innovation which uses an instant feedback system to interfaces the robot with its instructor’s brain. So essentially, the robot reads our thoughts and looks for a particular signal that our brain generates when we detect a mistake.
This signal is called “error-related potentials”—or ErrPs, for short, and is detected with the instructor’s electroencephalography (EEG) monitor placed on his/her head in real-time for data streaming. The ErrP signals are usually faint, so the software connected to Baxter has to be calibrated down to the last decimal. But this hard work is worth it since it makes it easier to train the robot and let it know when it’s performing an assigned task improperly. Whenever Baxter detects an ErrPs signal, it uses its feedback mechanism to make behavioral corrections in as little as ten milliseconds, which means the flow of the work continues without any significant interruptions.
The researchers need a lot more work to make the process mainstream, but they are confident that their system can achieve over 90 percent accuracy which would significantly reduce the training time for each robot. And even if it does screw up, the robot can just stop and ask for instructions rather than continuing with the error and possibly costing the company a whole lot of fortune.
What are your thoughts on this innovative solution? Comment below!