Control and learning for robot manipulators utilizing geometric frameworks
Vision-based robot policy learning
Adaptive, robust, learning, and nonlinear control.
Modeling, simulation, and management of traffic network systems, intelligent vehicle highway highway systems (IVHS)
Mechatronic systems, micro-electromechanical systems (MEMS)
Robust and data-driven controller design of multi-actuator hard disk drives (HDD)
We are developing a new end-to-end learning framework for autonomous robotic assembly and construction tasks that require contact-rich interactions by learning from a small number of expert demonstrations. This will be accomplished by extending the application scope and capabilities of our recently developed Equivariant Descriptor Fields (EDFs) machine learning models [4,7]: from learning end-to-end pick and place skills utilizing visual perception, to learning end-to-end compliant assembly skills that encompass motion and force contact interaction primitives utilizing visual perception, internal and external force/torque measurements, and tactile sensing. Our initial task will be to seamlessly integrate recently developed bi-equivariant diffusion- EDFs’ [4] vision-based task planning models with low-level equivariant geometric learning impedance controllers [1,2,3,5,6]. Full SE (3) equivariance in both vision and control forces can be achieved by combining these two approaches, resulting in efficient and robust learning of visual coordinate and force interaction skills to perform manipulation tasks in contact-rich environments.
[1] Nikhil Potu Surya Prakash, Joohwan Seo, Koushil Sreenath, Jongeun Choi, and Roberto Horowitz. "Deep Geometric Potential Functions for Tracking on Manifolds." IROS 2024, Abu-Dhabi, UAE [Link]
[2] Nikhil Potu Surya Prakash, Joohwan Seo, Koushil Sreenath, Jongeun Choi, and Roberto Horowitz. "Variable Impedance Control using Deep Geometric Potential Fields." MECC 2024, Chicago, USA [Link]
[3] Joohwan Seo, Nikhil Potu Surya Prakash, Jongeun Choi, and Roberto Horowitz. "A Comparison Between Lie Group-and Lie Algebra-based Potential Functions for Geometric Impedance Control." ACC 2024, Toronto, Canada [Link]
[4] Hyunwoo Ryu, Jiwoo Kim, Hyunsuk Ahn, Junwoo Chang, Joohwan Seo, Taehan Kim, Yubin Kim, Chaewon Hwang, Jongeun Choi, and Roberto Horowitz. "Diffusion-EDFs: Bi-equivariant Denoising Generative Modeling on SE(3) for Visual Robotic Manipulation." Highlight on CVPR 2024, Seattle, USA. [Link]
[5] J. Seo, N. P. Prakash, X. Zhang, C. Wang, J. Choi, M. Tomizuka, and R. Horowitz. “Contact-rich SE(3)-equivariant Robot Manipulation Task Learning via Geometric Impedance Control.” IEEE Robotics and Automation Letters, 2023. [Link]
[6] Joohwan Seo, Nikhil Potu Surya Prakash, Alexander Rose, Jongeun Choi and Roberto Horowitz. “Geometric Impedance Control on SE (3) for Robotic Manipulators.” IFAC World Congress 2023, Yokohama, Japan July 2023. IFAC-PapersOnLine, 56(2):276–283, 2023. [Link]
[7] Jiwoo Kim, Hyunwoo Ryu, Jongeun Choi, Joohwan Seo, Nikhil Potu Surya Prakash, Ruolin Li, and Roberto Horowitz. “Robotic Manipulation Learning with Equivariant Descriptor Fields: Generative Modeling, Bi-equivariance, Steerability, and Locality,” Oral presentation, Workshop on Symmetries in Robot Learning at Robotics Science and Systems (RSS). July 2023. [Link]