If you have an engineering degree, you might want to skip to this paper published about the beer-fetching robots at MIT. It will probably explain it in actual robot terms, as opposed to beer terms, in which I’m going to break it down right now.
A team of MIT researchers has put together a system of beer coolers on wheels that manage to act as waiters in order to deliver a payload — in this case, beer — to multiple locations at once. That may sound fairly uncomplicated, but not only are you teaching a robot to pick up beers and dump them in to a waiting cooler-bot (they call them Turtlebots because they’re named after Ninja Turtles, or famous painters, take your pick), but you’re teaching the server robots to take orders from multiple locations, and then check when the bar is ready to serve them a beer.
Instead of having a fixed set of instructions, in other words, the robots have to decide for themselves what they should be doing, without knowing exactly what the other robots in the group are doing. In this example, they’re bringing beers to the engineers. Obviously, you’re going to use this system mostly for beer, but, when fine-tuned, it could be applied to other real-world situations, like for work in a warehouse.
MIT put together a demonstration video, and it has no music or sound effects or narration because they were too busy drinking their Yuengling:
In the demo video, you can see that, when no one wants a beer, the Turtlebot takes it to the next room that ordered a beer, instead of, say, running itself in to a wall or something. The hitch here is that the robots can’t really communicate with each other unless they’re in close proximity. That design is based on the fact that, in the real world, WiFi is kind of unreliable. So, what they’ve really done is programmed a bunch of little robots to work within a set of directions, and to gather as much information as they can to assign themselves their next task, without having a programmer tell them what they should do next.
It’s all pretty rad, in the awesome sense.
(Via IEEE Spectrum)