Robots that watch for social cues could feed people by gauging when they are ready for a mouthful. This may make it easier for people who can’t feed themselves, such as those with tetraplegia, to socialise.
People who can’t control their legs or arms can use commercial robotic arms to help them eat. These use set time intervals between mouthfuls or manual triggers, but they can be awkward in social settings and interrupt conversation. Now at Cornell University in Ithaca, New York, and his colleagues have trained an artificial intelligence to pick up the social cues at meals via videos of 30 groups of three people eating together in a social setting.
The AI model learned to recognise the behaviour that people – both the person taking a mouthful and their companions – displayed immediately before the act of eating, which the researchers say in their paper is a “delicate dance of multimodal signalling”.
Advertisement
Their model, called Social Nibbling Network (SoNNET), could be used to tell a robotic arm when to feed a mouthful to a person who is paralysed, while avoiding any interruption to the social element of the meal.
To put the model to the test, the team recruited 10 groups of three diners for trials with a commercially available robot arm, with one person in each group being selected to constrain their arms and be fed by the robot. The participants were all students who don’t have a disability, because of the practical and health difficulties of gathering people with movement problems for group tests during a covid-19 lockdown.
The experiments used one of three strategies to initiate the robot to supply a mouthful of food: a set time interval; feeding a mouthful when the diner opened their mouth and kept it open as a cue; and SoNNET. The participants subsequently rated SoNNET as the best option in questionnaires.
at UK spinal injury charity Aspire says that paralysis has a catastrophic and devastating effect on people, and any recovery of a sense of independence can be enormously beneficial.
“To give them that freedom and independence to undertake a normal daily activity is very powerful,” he says.
Carlin says that although affordable voice recognition technology and smart assistants have had a dramatic positive effect on people who are paralysed, expensive technologies with more advanced benefits, such as exoskeletons and feeding robots, are out of the price range for most people.
“It’s something that, for those who can afford it, could be really, truly transformative. It shows that artificial intelligence can do some pretty amazing things to support a human being who has lost functionality, and it can give them independence back. So I think it’s brilliant,” says Carlin.
Reference:
