Visual Positioning-Based Disturbance Rejection Control for Power Grid Inspection Quadrotors

Aiming at the difficult problem of GPS-denied positioning for power grid inspection drones in complex electromagnetic environments, this paper proposes a quadrotor disturbance rejection control architecture that integrates improved visual recognition and Active Disturbance Rejection Control (ADRC). A target detection framework is constructed based on the lightweight YOLOv8 model. Visual positioning is achieved by integrating a pose calculation algorithm. A double-loop ADRC controller is designed, and the visual positioning deviation is dynamically injected into the Extended State Observer (ESO) to achieve online estimation and dynamic compensation of composite disturbances. This study provides a verifiable simulation methodology framework for the high-precision control of drones in GNSS-restricted environments.