Background: AI has shown the potential to positively influence minimally invasive surgical robotics. Incorporation of AI can improve perception, planning, decision-making, and execution in this field. Existing systems include, but are not limited to, the STAR for supervised autonomous suturing, MAKO for orthopedic arthroplasty, CyberKnife for radiosurgery, and PROST for prostate interventions. Such systems have shown advancement in precision, image guidance, task automation, haptic feedback, and flexible safety management. However, unlike domains such as autonomous driving, surgical robotics has progressed more cautiously. Current platforms have been found to sporadically lack transparent supervision contracts, adequate surgeon-centric safety guarantees, standardized pathways for adaptive autonomy, embedded safeguards within their modes of operation, and validated metrics for assessing performance. Objective: This paper presents a conceptual framework for a human-supervised AI-driven smart actuator system focused on minimally invasive surgery. The goal of this paper is to propose a forward-looking proof of concept that formalizes surgeon authority, integrates AI-enabled perception and control, enforces provable safety constraints, enables adaptive assistance, and ensures continuous, patient-safe force regulation. Conceptual Design: This architecture incorporates compact backdrivable actuators. It also includes multimodal sensing that encompasses sensor data, force and torque, pose, endoscopic vision, and tissue impedance, as well as an AI stack based off a machine learning and reinforcement learning framework. The model delineates three operational modes. These include teleoperation enhanced by AI-based overlays, shared control that incorporates tremor suppression, virtual fixtures, as well as force regulation, and super-vised autonomy where specific subtasks are carried out under surgeon pedal-hold and confidence gating. Safety is ensured using control barrier functions and model predictive safety filters, which block unsafe actions applying reinforcement learning. Aside from this, human-factor elements feature confidence-aware visualization, multimodal anomaly detection, and options for immediate overrides. Contribution: This study outlines a research roadmap. Our contributions include a formalized supervised-autonomy contract, a layered safety design combining reinforcement learning with provable constraint enforcement, surgeon-centered framework with immediate veto and transparency features, and a translational agenda spanning simulation, phantom, ex-vivo, and cadaveric validation. Conclusion: This paper aims to position AI as a cooperative assistant rather than an autonomous decision-maker in the field of robotic and precision surgery. The conceptual framework endeavors to address challenges in surgical robotics. These include precision, safety, transparency, and supervisory oversight. It synthesizes lessons from current exemplars. It articulates a pathway toward adaptive, auditable, transparently supervised, resilient, and patient-centered surgical systems.
Saxena et al. (Fri,) studied this question.