There's a long list of reasons why you don't see a lot of non-vacuum robots in homes. First and foremost are the issues of unstructured and semi-structured environments. From layout, lighting, and surfaces to people and pets, no two homes are the same. Even though robots can effectively map every home, space is always changing.
This week, researchers at MIT CSAIL demonstrated a new way to simulate and train home robots. Using an iPhone, people can scan a portion of their home and then upload it to a simulated environment.
In recent decades, simulation has become a fundamental element of robot training. It allows robots to attempt thousands or even millions of failed tasks in the same amount of time it takes to complete one in the real world.
The consequences of failure in simulation are also significantly lower than in real life. Imagine if you were to teach a robot to put mugs in the dishwasher, it would have to break 100 real-life mugs in the process.
"Simulating training in a virtual world is very powerful because the robot can practice it millions of times," researcher Pulkit Agrawal said in a video accompanying the study. "It probably broke a thousand plates, but it didn't matter because everything was happening in the virtual world."
However, like robots themselves, simulation can only go so far when it comes to dynamic environments like the home. Making simulation as convenient as iPhone scanning can greatly improve the robot's ability to adapt to different environments.
In fact, building a strong enough environmental database will ultimately make the system more adaptable when inappropriate situations inevitably arise, such as moving furniture or placing dishes on the kitchen counter.