Robust Decentralized H∞ Attack-Tolerant Observer-Based Team Formation Network Control of Large-Scale Quadrotor UAVs: HJIE-Reinforcement Learning-Based Deep Neural Network Method

In this paper, a robust decentralized $H_{\infty} $ attack-tolerant observer-based team formation tracking control scheme is proposed for large-scale quadrotor unmanned aerial vehicle (UAV) systems under external disturbance, measurement noise, couplings from other neighboring quadrotor UAVs, and malicious attacks on actuator and sensor of the network control system (NCS) via wireless communication. First, we constructed a smoothed model of attack signals to describe their behavior. Then, by integrating the smoothed dynamic model with the system dynamic model of each quadrotor UAV, we can simultaneously estimate the attack signals and the system state of each quadrotor UAV through a traditional Luenberger observer for the efficient robust decentralized $H_{\infty} $ attack-tolerant observer-based team formation tracking control of large-scale quadrotor UAVs. For the design of robust decentralized $H_{\infty} $ attack-tolerant observer-based team formation tracking control of large-scale quadrotor UAVs, a very difficult independent nonlinear partial differential observer/controller-coupled Hamilton Jacobi Issac equation (HJIE) must be solved for the observer and controller design of each quadrotor UAV. Nowadays, there are no analytical and numerical methods to resolve HJIE. Thus, an HJIE-reinforcement learning-based deep neural network (DNN) is trained to directly solve the observer/controller-coupled HJIE for robust decentralized $H_{\infty} $ attack-tolerant observer-based team formation tracking control of each quadrotor UAV. Since the system model of the quadrotor UAV and HJIE have been adopted for the HJIE-reinforcement Adam learning algorithm DNN training, compared to the traditional DNN big data-driven training schemes, we save a lot of training data and time to achieve the robust decentralized $H_{\infty} $ attack-tolerant observer-based team formation tracking control design. As the Adam algorithm converges, we could show that the proposed HJIE-reinforcement DNN-based decentralized $H_{\infty} $ attack-tolerant observer-based tracking control scheme can achieve the theoretical result. Finally, the simulation results are presented with a comparison to verify the effectiveness of the proposed method.