Delivery robots have been popping up everywhere these days. Tests have been largely successful but these are usually done in smaller controlled areas. Once delivery bots start being used on a wider scale, they may have some issues. For instance, thanks to GPS, they can pretty easily find a house, but at some properties, the robot may have trouble finding the front door.
Normally the AI maps the area and building ahead of time using algorithms and/or GPS to help guide the robot. As delivery zones grow larger, it can become impractical to map every building in an entire city. Another issue is that exteriors of houses can often change with the seasons. Mapping someone’s house could also lead to privacy and security issues.
Delivery Robots Analyze Environment to Find Front Door
Engineers at MIT may have a solution for this that doesn’t require mapping an area and its buildings beforehand. Instead, the robot will use machine learning to recognize objects and analyze the environment in real-time to make a determination of the proper place to leave the package.
For instance, once it’s at the address, the AI will look for clues that will ultimately lead it to the front door. It might start on the road but then recognize a driveway which it has been trained to determine as likely leading to a sidewalk which will ultimately lead to the front door. This technique allows the AI to make a decision on its own of the best way to find the location rather than relying on pre-programmed coordinates.
Planning Beyond the Sensing Horizon Using a Learned Context
This technique would eliminate the need for relying on a map for a specific residence and cut down on time a robot spends exploring a property before finding the proper place to drop the package. Through machine learning, they can use context to efficiently locate a specific location, even in unfamiliar, unmapped environments.
“We wouldn’t want to have to make a map of every building that we’d need to visit,” says Michael Everett, a graduate student in MIT’s Department of Mechanical Engineering. “With this technique, we hope to drop a robot at the end of any driveway and have it find a door.”