Label-Free Adaptive Gaussian Sample Consensus for Learning from Perfect and Imperfect Demonstrations

Autonomous robotic surgery is one of the most groundbreaking advancements in medical technology. Learning from human demonstrations is promising in this domain, which facilitates the transfer of skills from humans to robots. However, the practical application of this approach is hampered by the difficulty of acquiring high-quality demonstrations since surgical tasks often involve complex manipulations and stringent precision requirements, leading to frequent operation errors in the demonstrations. These imperfect demonstrations adversely affect the performance of controller policies learned from the data. Unlike existing methods that rely on extensive human labelling of demonstrated trajectories, we present a novel, label-free iterative optimization approach that enables the continuous intake of imperfect demonstrations without the need for tedious labelling by introducing a label-free adaptive Gaussian sample consensus method, thereby progressively refining the control policy. We demonstrate the efficacy and practicality of our approach through two experimental studies: a handwriting classification task, providing reproducible ground-truth labels for evaluation, and an endoscopy scanning task, demonstrating the feasibility of our method in a real-world clinical context. Both experiments highlight our method's capacity to efficiently adapt to and learn from an ongoing stream of imperfect demonstrations.
See our lab website for more details and a copy of the paper: www.ece.ualbert... (go to Publications).