'Smart glove' could help robots get a (better) grip on objects
Robots could one day achieve greater precision and dexterity in food and object handling thanks to a new ‘smart glove’.
The glove — developed by US researchers — contains 548 sensors which feed data to a deep learning network. The network then uses this information to identify objects, estimate their weight and respond to tactile feedback.
To get the data, researchers wore the glove and recorded themselves handling a variety of objects, including a horned melon, tea box, coin, spoon and mug to create a tactile ‘map’ which allowed the network to recognise objects based on the way they were held.
The method takes advantage of humans’ proficiency in manipulating objects based on their weight, shape and hardness without dropping or breaking them — an ability which is difficult to engineer in robots.
While research into vision-based grasping strategies has progressed substantially, little data existed on the tactile information humans rely on when grasping objects.
However, the glove’s simplicity, scalability, ability to be used over long intervals and low cost has allowed the researchers to generate a large dataset that complements vision-based robot-object handling and could inform future robot design.
Unlocking the benefits of avocados across the value chain
Westfalia Fruit is helping to advance the future of the avocado sector with its zero-waste,...
Risk management approach for toxin in infant formula
Following recent recalls of specific infant formula products, the European Food Safety Authority...
Fast detection of food contamination using enhanced AI tool
A deep learning-based model has been developed to detect contaminants within three hours in foods...
