A Vision-Based Robust Adaptive Control for Caging a Flood Area via Multiple UAVs

This paper presents a control strategy for caging a flood area via multiple UAVs. The strategy consists of the following parts. A novel architecture for video segmentation, Multiscale Features Fusion based MobileNet (MFFM-Net), is constructed to detect the flood boundary. A Function Approximation Technique based Immersion and Invariance (FATII) tracking controller is employed to constrain a single UAV on the flood boundary in the presence of external disturbances. A flocking based formation controller is designed to uniformly distribute UAVs along the flood boundary without collisions among neighbours. The proposed strategy has been verified through simulations under the ROS/Gazebo environment.