An Adaptive Neural Network State Estimator for Quadrotor Unmanned Air Vehicle

Jiang Yuning,Qian Bo,Ghulam Farid,Sohaib Tahir,Muhammad Ahmad

doi:10.14569/ijacsa.2019.0100242

Abstract

An adaptive neural observer design is presented for the nonlinear quadrotor unmanned aerial vehicle (UAV). This proposed observer design is motivated by the practical quadrotor where the whole dynamical model of system is unavailable. In this paper, dynamics of the quadrotor UAV system and its state space model are discussed and a neural observer design, using a back propagation algorithm is presented. The steady state error is reduced with the neural network term in the estimator design and the transient performance of the system is improved. This proposed methodology reduces the number of sensors and weight of the quadrotor which results in the decrease of manufacturing cost. A Lyapunov-based stability analysis is utilized to prove the convergence of error to the neighborhood of zero. The performance and capabilities of the design procedure are demonstrated by the Simulation results.

Highlights

In recent years, quadrotor has become an interesting research area for the robotics community in the field of autonomous aerial vehicles
The problem of estimating system state has already been done using Kalman filter [3], but the most important problem with the design procedure of classic observers is the presence of external disturbance and unknown dynamical model
In [10], a new Neural Network Observer (NNO) is designed to estimate the translational and angular velocities of the unmanned aerial vehicle (UAV), and an output feedback control law is developed in which the position and the attitude of the UAV are considered as a state variable to control the aircraft more accurately

Summary

INTRODUCTION

Quadrotor has become an interesting research area for the robotics community in the field of autonomous aerial vehicles. In [10], a new Neural Network Observer (NNO) is designed to estimate the translational and angular velocities of the UAV, and an output feedback control law is developed in which the position and the attitude of the UAV are considered as a state variable to control the aircraft more accurately. In [11], a new dynamic neural network based observer is presented and is proved using sliding mode stability analysis so in the presence of uncertainty, disturbance and sensor noise it could asymptotically track the states of a quadrotor and blade flapping. The point is that this work has considered the whole dynamic of quadrotor model undefined and the result shows the capability of neural network in the prediction and estimation of nonlinear functions.

NEURAL NETWORK STATE ESTIMATOR

STABILITY PROOF

SIMULATION RESULTS

CONCLUSION