Rearranging and manipulating deformable objects such as cables, fabrics, and bags is a long-standing challenge in robotic manipulation. The complex dynamics and high-dimensional configuration spaces of deformables, compared to rigid objects, make manipulation difficult not only for multi-step planning, but even for goal specification. Goals cannot be as easily specified as rigid object poses, and may involve complex relative spatial relations such as “place the item inside the bag”. In this talk, we present three techniques that take a step towards addressing challenges in deformable object manipulation by using model-free imitation learning, model-based reinforcement learning, and action-centric architectures known as Transporter Networks.
Daniel Seita ’14 is pursuing a PhD in computer science at the University of California, Berkeley, and earlier received Bachelor’s degrees in Computer Science and Mathematics from Williams College. His research focuses on robotic manipulation, imitation learning, and reinforcement learning, particularly in the contexts of deformable object manipulation and surgical robotics. Daniel’s research is supported by a six-year Graduate Fellowships for STEM Diversity, and was previously supported by a two-year Berkeley Fellowship. He is the recipient of the Honorable Mention for Best Paper award at UAI 2017 and the 2019 Eugene L Lawler Prize from the Berkley EECS department.