No one enjoys picking up after themselves, and in the future, you may not have to, because your robot butler will be cleaning your house for you, always ready and able to lend a helping hand. That’s the world researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the University of Toronto are envisioning.
Their new paper details how they are training robots, called artificial agents, to learn complex commands within household chores through a 3-D world inspired by “The Sims” video game.
Robots are being taught to do our tedious chores
One reason why we don’t all have robot butlers at our beck and call is that it is hard to teach a robot to infer and reason like we can. Take the command “turn on the TV.” We would know that this means we need to fetch the remote and use the power button on it. We also know that we would need to be in range of the TV to do this. A robot would get stumped on these implicit details.
To better teach a robot to pick up on these cues, the researchers used “VirtualHome,” a system which simulates household tasks in a furnished house that looks like our own. The artificial agent would watch the videos within VirtualHome to break down complex tasks into manageable ones. So far, the researchers have successfully put its artificial agent through 1,000 tasks like “making a cup of coffee.”
In the future, the researchers hope to teach its robot beyond the world of VirtualHome, and hope that it can learn by just watching a YouTube video. “There are many exciting avenues going forward, for example, training agents to perform tasks from visual observation alone,” the paper concludes.
Robots could one day learn by observing us instead of having our commands exhaustively explained to them. For those of us who need help managing the day-to-day demands of a household, its help would be a welcome relief. “You can imagine a setting where robots are assisting with chores at home and can eventually anticipate personalized wants and needs, or impending action,” the paper’s lead author, MIT PhD student Xavier Puig, said. “This could be especially helpful as an assistive technology for the elderly, or those who may have limited mobility.