Online Adaptive Dynamic Programming for Optimal Position Control of Magnetically Actuated Capsule

In this article, an optimal position control framework based on online adaptive dynamic programming is proposed for magnetically actuated capsules (MACs). First, sensorless magnetic force optimization is proposed by treating the magnetic force as an external force included in the manipulator torque, such that the implicit relationship between the manipulator torque and the image data is constructed to eliminate the need for system identification and additional sensor measurements. Second, a total disturbance local dynamic linearization technique is designed to ensure the positioning stability of the MAC through decoupling total disturbances (e.g., friction and parameter estimation errors) without the use of any disturbance observers. Subsequently, an optimal position control framework is constructed using a time-based Hamilton–Jacobi–Bellman equation to eliminate the uncontrollable iterative process within the gastrointestinal system. Finally, three scenarios with the internal camera are conducted in broad and narrow areas to validate the effectiveness of the proposed optimal control scheme.