Whole-Body Model Predictive Control for Magnetically Actuated Robotic Endoscope

Magnetically actuated robotic devices are gaining increasing attention in medical applications. Robotic devices based on permanent magnets are better suited for the human body due to their wide workspace for operation. For these devices, handling the motion constraints of external permanent magnets (EPM) reasonably is crucial for safety. These constraints include motion range and obstacles, and the internal permanent magnet (IPM) motion needs to be controlled simultaneously. Existing hierarchical control structures usually design the controllers for the IPM and the manipulator separately. This makes it difficult to simultaneously satisfy the movement constraints of the EPM and the control accuracy of the IPM. To address this problem, we model the manipulator and IPM as a whole body and yield control command by formulating an optimization problem. We simplify the model and formulate a real-time iteration model predictive controller (RTI-MPC) to improve computational efficiency, and the control frequency reaches 30 Hz in actual deployment. We have validated the feasibility and tested the performance of this method through simulations and experiments. The experiments demonstrate that the manipulator can successfully control the IPM through the colon phantom while avoiding obstacles in the task space.