Sliding Mode Control and Adaptive Control are widely studied in the area of Rotor UAV in recent years. Although the performance of Rotor UAV with these controllers show high command tracking ability and good robustness, they are limited by model accuracy so that they cannot take place PID. In this paper, a novel method named State damping control is proposed to be a candidate for the traditional PID method. Our proposed State Damping Control is inspired by the format of air resistance. The method is based on the general idea that resistance will make a system easy to stabilize. State damping control is independent of model accuracy and just uses three parameters to control attitude, so it is easy to realize. Krasovskii Theorem is used to give the evidence that State damping control is asymptotic stable in our considered state space. Finally, simulations are implemented in C++ on VS2017, it demonstrates that State damping control is easy to be tuned and robust to wind attack and inertial parameters. Compared with PID, our proposed method is robust to wind disturbances obviously.

Forest fires can cause tremendous damage to the environment and irreparable loss to individual properties. Using deep learning algorithms on embedded sensing devices to detect smoke and flame for real-time detection of forest fire is essential to minimize the risk to the environment and individual properties. This paper proposes an ultra-lightweight network, YOLO-ULNet, for real-time detection of forest fire on embedded sensing devices. Initially, a lightweight network, YOLO-LNet, consisting of input, backbone, neck, and output head was designed to reduce the parameter quantity and computation quantity. The validation results of YOLO-LNet on the computer showed that the mAP@0.5, speed, computation quantity, and parameter quantity were 71.2%, 110 FPS, 0.7 GFLOPs, 0.26 M, respectively, representing a performance that was superior to that of YOLOv5s. The test results of YOLO-LNet on the Raspberry Pi (RPi) 4B showed that the precision and speed were 77.75% and 14.39 FPS, respectively, where the speed required for real-time detection (20 FPS) still could not be achieved. Subsequently, channel pruning and feature distillation model compression methods were employed to obtain the ultra-lightweight network YOLO-ULNet with improved speed and precision. The validation results of YOLO-ULNet on the computer showed that the mAP@0.5, speed, computation quantity, and parameter quantity were 69.1%, 115 FPS, 0.4 GFLOPs, and 0.19 M, respectively. Meanwhile, the test results of YOLO-ULNet on the RPi 4B showed that the precision and speed were 74.50% and 24.57 FPS, respectively, meeting the requirements of the real-time detection of forest fire.

Tracking trajectory with high precision under wind perturbations is a difficult problem of a quadrotor. In response to this problem, a novel wind perturbation estimator using neural network is proposed. The structure of wind estimator is designed based on deep understanding of a quadrotor, and training process is optimized to solve overfitting problem in the presence of sensor noise. With the consideration of wind perturbations and rotor dynamics, cascaded Lyapunov functions are used to derive full backstepping controller. Compared with traditional wind estimator, the proposed estimator is simple to be carried out and shows better robustness to sensor noise. To the best of our knowledge, the proposed controller is more robust to wind with less power cost than existing controllers. A series of simulations show the process to optimize wind estimator, and comparison between different controllers demonstrates that the proposed controller is robust to wind and energy-efficient. Finally, experiments have strengthened the effectiveness of our proposed method.