The way that robots learn is much different from the way that humans learn. People are able to learn by observation. You can teach a human how to do something simply by showing them. If you want to teach someone how to make a salad, you show them how. They will learn how to slice the tomatoes, and chop the cucumber, and wash the greens. They will recognize that you have to toss the salad and try and evenly distribute all of the ingredients.
However, industrial robots have always had to be programmed to carry out tasks. If you have a robot that isn’t programmed to make a salad, it won’t be able to make a salad. You can’t just say, “Look here robot. I’m only going through this once, so pay attention.” Robots don’t learn from watching people perform actions. They must be programmed to carry out tasks.
Well, that’s how things have worked up until recently. Researchers at the University of Maryland are hoping to change the way that robots learn. They want robots to be able to learn how to carry out complex tasks by observing humans rather than having to rely on programming.
The researchers have already found success with this approach, having trained a robot to mix a cocktail by watching an expert mix a drink. While the idea of a robotic bartender is appealing, it’s not going to dramatically improve your life. However, teaching robots to learn through observation could prove to benefit manufacturing.
Teaching robots to learn through observation rather than programming provides some significant benefits to the manufacturing industry. Robots that learn by watching humans could prove to be more versatile. Instead of being programmed for one specific task, robots that learn through observation could carry out a number of tasks depending on what was required of them. Not only would this make robots more versatile, but it would also decrease the amount of time needed to retool manufacturing lines.