Fast Configuration Identification and Matching Gains of RFA Based on Deep Learning Technology

In order to expand the application scene of rotorcraft UAVs, a reconfigurable flight array(RFA) has been proposed, which can be quickly assembled into different configurations to adapt complex environments. However, the flight stability of RFA depends on the controller, which depends on the model, and the models of different configurations are inconsistent. In this paper, we first give a general dynamic model of reconfigurable flight array, which shows that obtaining the coordinates of each unit module is the key to design stable control schedule. Then, a target detection algorithm is employed to identify the current configuration and the unit modules as to find the coordinates of each unit module. Finally, the tracking controllers of the two configurations are matched according to the configuration information of visual recognition, and the module coordinates obtained by visual recognition and manual recognition are simulated and compared. The results show that the method proposed in this paper can identify the right configuration quickly and match suitable tracking controller for RFA to achieve fast and stable control.