The final step in a laundry pipeline is to fold garments for their later storage. Folding garments prevents them to get wrinkled and reduces the required storage space. Each garment has associated a series of folding operations or steps depending on its garment category (i.e. folding a t-shirt takes a different set of actions that folding some pants). Therefore, most of the existing works on folding garments focus on detecting the garment category from the extended garment to apply a predefined folding sequence.
Our current research, on the other hand is focused on learning how to fold garments from a human teacher without the need for a predefined folding sequence. For this we are using Continuous Goal-Directed Actions combined with Deep Reinforcement Learning.