UAV Attitude Prediction using a new EKF and Stochastic Disturbed Process Noise Model

Abstract

Even though low-cost inertial navigation units (INS) are affordable, their data gathering is often of low quality and accuracy. In addition, the nonlinear nature of process and measurement models renders those INS incapable of delivering accurate navigation for airborne vehicles. Given the constraint that MEMS sensors used in low-cost INS design offer signals with bias variations and large amounts of noise, the present paper aims to design an EKF for low-cost UAV attitude estimation using a perturbed process noise model and a nonlinear measurement estimator. The UAV orientations are obtained using a data fusion model that uses accelerometers and magnetometer measurements gathered from field experiments. First-order Gauss-Markov models and zero-mean Gaussian noises are used to model biases and measurement noises, respectively. Simulation results reveal that the proposed method can effectively estimate the UAV attitude with high precision and cheap cost, both of which have practical applications in flight control systems.