Indoor fixed-height control for bio-inspired flapping-wing aerial vehicles based on off-board monocular vision

The flapping-wing aerial vehicle (FWAV) is a new kind of aerial vehicle that imitates the flapping wings of birds or insects during flight and has the advantages of flexible flight, high flight efficiency, and good concealment compared with the fixed-wing and the rotary-wing aerial vehicles. Therefore, the FWAV has attracted considerable attention in recent years. However, the flight mechanism of the FWAV is complex and has many motion parameters with strong coupling. Thus, establishing a precise and practical motion model is difficult. At the same time, given the limited weight and load capacity of small FWAVs, it cannot carry accurate but heavy positioning equipment. Thus, many problems in autonomous flight control of FWAVs need to be addressed at this stage. For the fixed-height flight of FWAVs, an indoor fixed-height control system based on off-board monocular vision was designed. First, image sequences of the FWAV were obtained using the off-board monocular camera. Then, the ground station software based on Qt received the images, detected the light-emitting feature point on the FWAV, and obtained the pixel coordinates of the feature point on each image using the OpenCV image processing algorithms. On the basis of the Kalman filter, the image state estimator of the feature point was established to reduce the environmental interference and solve the temporal missing data problem of the feature point. Finally, the conventional and single-neuron PID control systems were established, where the motor speed of the FWAV was controlled by Bluetooth, to achieve image-based indoor fixed-height flight of the FWAV. Experimental results show that the fixed-height flight control system designed in this study can keep the image coordinates of the feature point of the FWAV at the center of the camera image. For the step signal, the response speed of the single-neuron PID control system is slightly slower than that of the conventional PID control system. However, the control accuracy of the single-neuron PID control system is better than that of the conventional PID controller, with a maximum relative error of 3%.