Adaptive Low-Computation Neural Network Control of Constrained Nonlinear Systems and Its Application to Agricultural Drones

This research presents an adaptive low-computation neural network (NN) controller for a kind of constrained nonlinear systems. Firstly, a linear state observer is constructed to estimate unmeasure states. To overcome the filtering error, a filtering-error compensation system is constructed. To deal with external disturbances and system non-linearity, an adaptive neural control strategy is developed. Especially, the number of RBFNN’s weight can be reduced greatly, which helps to decrease the computation pressure. By appliying barrier Lyapunov function, it proves that the tracking errors can drive into the bounded regions and full states can be kept within the preset constraints. Last but not least, a numerical simulation and a real example about the attitude control of agricultural drones are performed to verify the availability of the developed controller.