IPAB Workshop - 01/12/2022

Speaker: Xiaofeng Mao

Abstract: Tactile information is important in enabling robot to perform dexterous manipulation and grasping tasks. I will present a data-efficient learning from demonstration framework which exploits the use of rich tactile sensing and achieves fine dexterous bimanual grasping. Specifically, we formulated a convolutional autoencoder that can effectively extract and encode the contact information and features between the robot and the objects, based on the high-dimensional tactile feedback. Further, we developed a behavior cloning network that can learn human's sensorimotor skills demonstrated directly on the robot hardware and in the task-space, fusing both proprioceptive and tactile feedback. Our comparison study revealed the effectiveness of the rich contract information provided by tactile sensing in the demonstration data, which enabled successfully extraction and replication of the demonstrated motor skills.


Speaker: Jack Rome 

Abstract: Learning to manipulate cloth objects has applications in assistive care and domestic robotics. There is an increasing desire to create systems that can assist in tasks such as dressing and laundry. Most of us can complete these tasks with ease every day, however, some of us rely upon assistance from others. There is an opportunity to provide this assistance through robotics. Cloth manipulation tasks represent a variety of challenges for robots due to the complexity of complexity of cloth dynamics. Cloth will deform with high-resolution crumples and folds during manipulation; leading to infinite states of shapes, sizes, folds, wrinkles, and crumples on the same object. In addition to this, the variety of colours, patterns, shapes and sizes of garments make it difficult to generalise solutions. This presentation will detail these challenges and will survey current literature in cloth manipulation. Finally, this presentation will introduce my work so far in investigating these challenges in cloth manipulation and discuss my current work and plans for my PhD project."

Dec 01 2022 -

IPAB Workshop - 01/12/2022

Xiaofeng Mao/Jack Rome

G.07, IF