The unmanned aerial vehicle (UAV) trajectory tracking control algorithm based on deep reinforcement learning is generally inefficient for training in an unknown environment, and the convergence is unstable. Aiming at this situation, a Markov decision process (MDP) model for UAV trajectory tracking is established, and a state-compensated deep deterministic policy gradient (CDDPG) algorithm is proposed. An additional neural network (C-Net) whose input is compensation state and output is compensation action is added to the network model of a deep deterministic policy gradient (DDPG) algorithm to assist in network exploration training. It combined the action output of the DDPG network with compensated output of the C-Net as the output action to interact with the environment, enabling the UAV to rapidly track dynamic targets in the most accurate continuous and smooth way possible. In addition, random noise is added on the basis of the generated behavior to realize a certain range of exploration and make the action value estimation more accurate. The OpenAI Gym tool is used to verify the proposed method, and the simulation results show that: (1) The proposed method can significantly improve the training efficiency by adding a compensation network and effectively improve the accuracy and convergence stability; (2) Under the same computer configuration, the computational cost of the proposed algorithm is basically the same as that of the QAC algorithm (Actor-critic algorithm based on behavioral value Q) and the DDPG algorithm; (3) During the training process, with the same tracking accuracy, the learning efficiency is about 70% higher than that of QAC and DDPG; (4) During the simulation tracking experiment, under the same training time, the tracking error of the proposed method after stabilization is about 50% lower than that of QAC and DDPG.

When a mobile robot inspects tasks with complex requirements indoors, the traditional backstepping method cannot guarantee the accuracy of the trajectory, leading to problems such as the instrument not being inside the image and focus failure when the robot grabs the image with high zoom. In order to solve this problem, this paper proposes an adaptive backstepping method based on double Q-learning for tracking and controlling the trajectory of mobile robots. We design the incremental model-free algorithm of Double-Q learning, which can quickly learn to rectify the trajectory tracking controller gain online. For the controller gain rectification problem in non-uniform state space exploration, we propose an incremental active learning exploration algorithm that incorporates memory playback as well as experience playback mechanisms to achieve online fast learning and controller gain rectification for agents. To verify the feasibility of the algorithm, we perform algorithm verification on different types of trajectories in Gazebo and physical platforms. The results show that the adaptive trajectory tracking control algorithm can be used to rectify the mobile robot trajectory tracking controller’s gain. Compared with the Backstepping-Fractional-Older PID controller and Fuzzy-Backstepping controller, Double Q-backstepping has better robustness, generalization, real-time, and stronger anti-disturbance capability.

Tree branches near the electric power transmission lines are of great threat to the electricity supply. Nowadays, the tasks of clearing threatening tree branches are still mostly operated by hand and simple tools. Traditional structures of the multirotor aerial robot have the problem of fixed structure and limited performance, which affects the stability and efficiency of pruning operation. In this article, in order to obtain better environmental adaptability, an active deformable trees-pruning aerial robot is presented. The deformation of the aerial robot is implemented through two ways, arm telescopic and folding. In order to suppress the influence of internal and external disturbances on the system, Active Disturbance Rejection Control (ADRC) technology is adopted to build the flight controller. Firstly, active deformation aerial robot structure is given, followed by system dynamic model establishment under wind disturbance using the Newton–Euler method. Also, the analysis of the gusts influence on the system is considered. Then, the active deformation aerial robot system is decoupled into a combination of six SISO systems, so that a disturbance rejection controller is designed. Finally, the expanded state observer and the nonlinear state error feedback law are used to inspect and compensate the disturbance. Simulation results of attitude and position tracking as well as the antidisturbance capability show that the active deformation aerial robot with the ADRC flight controller designed in this paper has excellent attitude control capabilities during flight and trees-pruning operation.

This paper proposes a novel aerial manipulator with front cutting effector (AMFCE) to address the aerial physical interaction (APhI) problem. First, the system uncertainty and external disturbance during the system movement and contact operation are estimated by modeling the entire robot and contact position. Next, based on the established model, the nonlinear disturbance observer (NDO) is used to estimate and compensate the unknown external disturbance of the system and the uncertainty of the model parameters in real time. Then, the nonsingular terminal synovial membrane control method is used to suppress the part that is difficult to estimate. Finally, a controller which is suitable for the movement and operation of the entire system is designed. The controller’s performance is verified through experiments, and the results show that the design, modeling, and control of the entire system can achieve the APhI.