The 2025 IEEE International Conference on Robotics and Automation is in full swing! While no one from ARCO Lab is attending, our lab members are co-authors on two papers being presented. The first is Dr. Giamou’s work with Dr. Limoyo and Dr. Marić from STARS Lab on Generative Graphical Inverse Kinematics (GGIK) was published as a T-RO paper last year. This method leverages years of work on distance-geometric inverse kinematics to train a state-of-the-art deep network for inverse kinematics (IK) of redundant manipulators. Our architecture is unique in that a single network can be used to solve IK problems for multiple manipulator models with completely different geometric structures. GGIK’s performance is made possible by the distance-geometric view of IK, which lends itself to a graphical description and unlocks Euclidean equivariance for improved sample efficiency.
Our second paper is in collaboration with McMaster’s HeART Lab and concerns OMASTAR, a technique for designing magnetic actuation systems with convex optimization. Control of microrobots in surgical settings requires the application of external magnetic fields, but the presence of magnets generating these fields can obstruct surgeons’ access to patients. We present a method of designing a system of rotating permanent magnets housed beneath an operating table. The poses of magnetic actuators are chosen by the OMASTAR algorithm so as to maximize the manipulability of the system, which quantifies how much actuator effort is required to achieve a desired magnetic field. Congratulations to lead author Veerash Palanichamy and ARCO Lab’s own Hussein Saad for a job well done!
