Robots are already capable of doing many things for us, including cooking and serving our food. Certain delicate tasks can be quite tricky for them though, such as handling soft fruit. This video shows how a robot can be trained with a machine learning algorithm to peel a banana without crushing it.
As the researchers explain, this robot uses a “goal-conditioned dual-action deep imitation learning (DIL) which can learn dexterous manipulation skills using human demonstration data.”
This method:
predicts reactive action when the precise manipulation of the target object is required (local action) and generates the entire trajectory when the precise manipulation is not required.
[HT: Heecheol Kim]