If you’re visiting a friend and you find yourself stumbling sleepy at home in the middle of the night, basic common sense tells you that if you see a refrigerator, you’re probably in their kitchen. But a bathroom? More than likely you are in a bathroom. It is common sense approach to navigation that robots can finally copy to improve them on quickly finding your way around a house.
Carnegie Mellon University researchers working with Facebook’s AI research team have developed a new approach to Autonomous robot navigation system known as SemExp (cumbersome, but less cumbersome than Goal-Oriented Semantic Exploration) that uses deep learning not just to train them to recognize objects but also to know where in a house they are typically located.
Training a robot to move around your house is not impossible, but at the moment, requires a lot of manual input working together with image and object recognition so that the automaton knows which areas it has designated as a bedroom or kitchen, for example. But houses are constantly evolving, things are left everywhere, and objects, like a footstool trusted by a robot as a recognizable milestone in navigation, could be obscured by a pair of trousers abandoned someday.
SemExp’s goal is to make robots more flexible by giving them a basic form of common sense. Machine learning allows them not only to recognize individual objects, such as a coffee table in front of a kitchen table, or a dryer in front of a dishwasher, but also where in a user Those objects are more likely to be found at home. Sure, some lucky owners You could have a conveniently installed toilet in your office for maximum productivity. yesut for the most part, You will always find the large metal box called a fridge in the kitchen, or the oversized glossy panel on the wall in the family room.
One of the ways robot vacuums have dramatically improved performance and runtime is by intelligently navigating a house to avoid cleaning the same areas twice, or heading directly to a room that has not been recently tidied up. This research has similar goals, and the researchers hope they can make robots better at navigating a house so they don’t rely on specific objects that are in specific places all the time.
G / O Media may receive a commission
The improved adaptability also promises to make robots easier to interact with. If you ask a robot to bring you a cold drink from the fridge, you’ll know that the kitchen is the most likely place you’ll find that appliance, and it will do its best to head straight there first. But even your autonomous route planning will be improved as a result of having a better understanding of the objects around the house. If the robot sees a dining table, it could extrapolate that it is currently near the dining room, which is generally adjacent to the kitchen, allowing it to plot a shorter overall path. Most importantly, this research could one day make robots the ultimate tool for finding where you left your keys, or who stole the TV’s remote control.
.
