Neuro-adaptive fast integral terminal sliding mode control design with variable gain robust exact differentiator for under-actuated quadcopter UAV

In this paper, a robust global fast terminal attractor based full flight trajectory tracking control law has been developed for the available regular form which is operated under matched uncertainties. Based on the hierarchical control principle, the aforesaid model is first subdivided into two subsystems, i.e., a fully-actuated subsystem and an under-actuated subsystem. In other words, the under-actuated subsystem is further transformed into a regular form whereby the under-actuated characteristics are decoupled in terms of control inputs. In the proposed design, the nonlinear drift terms, which certainly varies in full flight, are estimated via functional link neural networks to improve the performance of the controller in full flight. Besides, a variable gain robust exact differentiator (VG-RED) is designed to provide us with estimated flight velocities. It has consequently reduced the noise in system’s velocities and has mapped this controller as a practical one. The finite-time sliding mode enforcement and the states’ convergence are shown, for all flight loops, i.e., forward flight and backward flight, via the Lyapunov approach. All these claims are verified via numerical simulations and experimental implementation of the quadcopter system in a Matlab environment. For a more impressive presentation, the developed simulation results are compared with standard literature. •Firstly, we are going to highlight that we have developed for the first time a regular form for the quadcopter system which is, quite confidently, a significant contribution. This regular form facilitates all the control strategies to be implemented on the under study system.•Secondly, we have defined a novel integro-differential sliding surface which provide us a novel non-singular control input for each sub-system of quadcopter. This strategy, on one hand, provides fast finite time convergence. On the other hand, this methodology also provides robustness against matched uncertainties.•Thirdly, a variable gain robust exact differentiator (VG-RED) has been utilized for the velocities estimation of each DOF. These estimations are very robust precise and almost noise free. These features certainly enhance overall robustness and practicability of the proposed strategy.•Fourthly, we have used feed forward neural network (FFNN) for the estimation of nonlinear unknown drift terms. The use of these FFNNs is one more step to the utilization of the proposed strategy in practical scenarios. So