Long-Range Navigation of a Transluminal Magnetic Miniature Robot via Mobile Actuation System

This paper presents the long-range navigation control of a helical magnetic miniature robot within 3-D transluminal environment through a sim-to-real interactive navigation scheme, which is realized by a mobile magnetic actuation system with dual adjustable electromagnetic coils and a human-robot interface. This method provides precise trajectory tracking control of the miniature robot by dynamic electromagnetic field regulation. In particular, the proposed system employs real-time detection transformer (RT-DETR) algorithm combined with a lightweight MobileNetV4 backbone network for real-time miniature robot detection, while enhancing trajectory smoothness through multi-frame association and Extended Kalman Filter (EKF) prediction. Meanwhile, electromagnetic coil currents are adjusted using predicted positions, PID control, and electromagnetic field pseudo-inverse methods to regulate the miniature robot motion. Experimental results demonstrate a detection accuracy of 98.7% in sharp-turn and tortuous paths, with average recognition and processing latency of 7 ms per frame, and trajectory tracking error of approximately 0.7 mm. The system successfully navigated a 730 mm path in complex vascular phantoms, which validated its robustness under occlusion and interference, demonstrating the feasibility and effectiveness of the proposed system.