DP4AuSu: Autonomous Surgical Framework for Suturing Manipulation Using Diffusion Policy With Dynamic Time Wrapping-Based Locally Weighted Regression

Background: Emerging imitation learning (IL) approaches have provided innovative solutions for completing surgical robotic suturing autonomously, significantly aiding surgeons in their manipulations. Methods: We introduce Diffusion Policy for Autonomous Suturing (DP4AuSu), a novel framework that leverages diffusion policy (DP) and dynamic time wrapping-based locally weighted regression to achieve autonomous robotic suturing. Results: In simulation, DP4AuSu achieved a 94% success rate for insertion subtasks over 50 trials. In a real-world setting, it achieves 85% success rate over 20 trials for suturing manipulations in 390.55–41.59s faster than conventional diffusion policy. Conclusions: Our novel framework can capture the multimodality in demonstrations and successfully learn the suturing policy and reduce the suturing time. To the best of our knowledge, this work represents the first application of diffusion policy for robotic suturing. We hope this research paves the way for the automation of more complex surgical tasks.