March 12 : Researchers at the University of Washington are developing a robotic system that can help make the process of feeding easier. After identifying the different kinds of food on a person’s plate, the robot can now strategize how to use a fork to pick up and deliver the food right to the individual’s mouth.
The researchers arranged plates that included a dozen different kinds of food, ranging in consistency from hard carrots to soft bananas. The plates also included different foods like tomatoes and grapes which generally have a tough outer layer and soft insides. Later the team provided the volunteers with a fork and asked them to pick up different pieces of food and feed them to a, mannequin. The fork had an in-built sensor t measure how much force people used when they picked up food.
The individuals used different strategies to pick up food with different consistencies. For eg, some volunteers skewered softer foods like bananas at an angle to stop them from slipping off the fork and for food like grapes and carrots, the individuals tried to use wiggling motions to increase the force and spear each crunch.
The robot, however, used the same force and skewering strategies strategy to try to pick up all the pieces of food, regardless of their consistency. It easily picked up hard foods but found difficulties with soft foods and those with tough skins and softer insides. Like humans, robots needed to adjust how much force and angle they used to pick up different kinds of food.
To create a proper skewering and feeding strategy that changed based on the food item, the researchers combined 2 different algorithms. First by using an object-detection algorithm called RetinaNet, which scans the plate and identifies the types of food on it and places a frame around each item.
Next, they developed SPNet, an algorithm that examines the kind of food in a specific frame and tells the robot the best way to pick up the food. For eg, SPNet tells the robot to skewer a strawberry or a banana slice in the middle and spear carrots at any one of the 2 ends.
The team then let the robot pick up pieces of food and feed them to the individuals using SPNet or a more uniform strategy, an approach that skewered the center of each food item regardless of its shape or size. SPNet differently stylized strategies either outperformed or performed the same as the uniform approach for all the kinds of food.
Source : http://www.washington.edu/news/2019/03/11/how-to-train-your-robot-to-feed-you-dinner/