همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Energy-Elastic Proximal Policy Optimization for Autonomous Soft Tissue Cutting in Robotic Surgery

نویسندگان :

Seyed Mohammad Mahdi Pazhouh¹ Ali Eslami² Ahmad ali Abin³ Hasan Ali Ahmadi⁴ Armin Salimi-Badr⁵

1- دانشگاه شهید بهشتی 2- دانشگاه شهید بهشتی 3- دانشگاه شهید بهشتی 4- دانشگاه علوم پزشکی شهید بهشتی 5- دانشگاه شهید بهشتی

کلمات کلیدی :

Robotic surgery،deep reinforcement learning،soft tissue cutting،energy-efficient tensioning،autonomous surgery

چکیده :

With the growing use of robotic systems in surgery, the need for autonomous and efficient soft tissue manipulation is increasing. This paper proposes an energy- efficient deep reinforcement learning method for autonomous soft tissue cutting. To narrow the simulation-to-reality gap, we vary spring stiffness coefficients to better capture the mechanical behavior of different tissues. We also introduce a pinch-point sampling strategy inspired by k-means++ to improve the spatial distribution of tensioning points, and a reward shaping term that penalizes unnecessary energy use to encourage smoother, more stable tensioning. Evaluated on 16 standard cutting patterns, our approach maintains cutting accuracy comparable to state-of-the-art methods while achieving higher stability and consistency. These results suggest that the proposed Energy-Elastic Proximal Policy Optimization framework can improve both the efficiency and reliability of robotic soft tissue manipulation.