Neural-Network-Based Adaptive Control of Wing Rock Motion

The wing rock motion is mathematically described by a nonlinear differential equation with coefficients varying with angle of attack. In this paper, a neural-network-based adaptive control system is developed for the wing rock motion control. The adaptive controller comprises a neural network controller and a compensation controller. The neural network controller using a recurrent neural network is utilized to approximate a nonlinear system dynamic function and the compensation controller is a compensator for the difference between the system dynamic function and the recurrent neural network. Since the recurrent neural network contains the dynamic characteristic, it has superior capabilities than feedforward neural network for the dynamic response. Thus, it is utilized as a dynamic approximator. Moreover, the Taylor linearization technique is employed to increase the learning ability of the recurrent neural network. The on-line parameter adaptation laws are derived based on a Lyapunov function; thus the stability of the system can be guaranteed. Simulations are performed to illustrate the effectiveness of the proposed neural-network-based adaptive control system. Simulation results demonstrate that the proposed design method can achieve favorable control performances for the wing rock motion control with completely unknown of the system dynamic function. NOMENCLATURE α = angle of attack φ = roll angle b = chord ρ = density of air l C = roll moment coefficient xx I = mass moment of inertia ∞ U = freestream velocity S = wing reference area