Intelligent Robotic Assistance in Dental Surgery: A Reinforcement Learning-Based Model

Robotics in dentoalveolar and oral maxillofacial surgery could allow for improved precision, predictability, and safety in difficult procedures. We introduce an RL-based method for robot intelligent control in dental surgery to enhance intraoperative decision-making and instrument handling. The system enables (semi-)automatic steering of surgical instruments by combining sensor information in real-time with 3D representation of the surgical space. The RL agent is trained in a v-rep virtual environment with reward signals for reduced tissue damage, optimized paths, and time-efficient execution. Experimental validation on synthetic and phantom models indicates the capability of the model to accommodate various anatomical discrepancies and to achieve accurate manipulation, surpassing the conventional rule-based robotic systems. This work demonstrates the feasibility of RL-guided robotic systems in dental surgery and paves the path for real-time clinical integration in the future.