An Intelligent Force-Sensing Method for Soft Parallel Robots in Gastrointestinal Endoscopy

Reliable force sensing remains a key challenge in robotic colonoscopy. This paper introduces an intelligent force sensing approach for a novel soft parallel robot with a crosss-tructured two-layer scaffold, which improves both axial and lateral stiffness and minimizes deformation interference. Force estimation combines the physical models of hydraulic actuator and robot, as well as neural networks, to address nonlinearities and friction, relying on actuation-side sensing. Experiments show high-fidelity force estimation across three axes, with root-mean-square error reaching about 27, 23, and 80 mN for x, y, and z-axes, respectively. This approach, compared to existing robotic systems for endoscopic submucosal dissection (ESD), eliminates the need for distal force sensors and complex integration at the instrument tip and offers lower system complexity and cost, while enabling reliable three-degree-of-freedom force sensing. It provides a practical and scalable solution for advanced force sensing in ESD and other endoluminal interventions.