Safety-Ensured Robotic Control Framework for Cutting Task Automation in Endoscopic Submucosal Dissection

Abstract: There is growing interest in automating surgical tasks using robotic systems, such as endoscopy for treating gastrointestinal (GI) cancer. However, previous studies have primarily focused on detecting and analyzing objects or robots, with limited attention to ensuring safety, which is critical for clinical applications, where accidents can be caused by unsafe robot motions. In this study, we propose a new control framework that can formally ensure the safety of automating the cutting task in endoscopic submucosal dissection (ESD), a representative endoscopic surgical method for the treatment of early GI cancer, by using an endoscopic robot. The proposed framework utilizes Control Barrier Functions (CBFs) to accurately identify the boundaries of individual tumors, even in close proximity within the GI tract, ensuring precise treatment and removal while preserving the surrounding normal tissue. Additionally, by adopting a model-free control scheme, safety assurance is made possible even in endoscopic robotic systems where dynamic modeling is challenging. We demonstrate the proposed framework in a simulation-based experimental environment, where the tumors to be removed are close to each other, and show that the safety constraints are enforced. We show that the model-free CBF-based controlled robot eliminates one tumor completely without damaging it, while not invading another nearby tumor.