UAV Attitude Estimation Using Nonlinear Filtering and Low-Cost Mems Sensors

Integration of low-cost Micro-Electro-Mechanical Systems (MEMS) in the design of Inertial Navigation Systems (INS) for real-time guidance, navigation, and control of flying vehicles often results in low precision, poor accuracy, and degraded performance. Most of existing filtering and data fusion methods that have been applied to the problem of attitude estimation with use of low-cost MEMS-based INS fail in cases that have highly nonlinear process dynamics, nonlinear measurement models, and long-range navigation. This paper presents two nonlinear filtering algorithms for low-cost Unmanned Aerial Vehicles (UAVs) attitude and measurement biases estimation. An Unscented Kalman Filter (UKF) and a Minimum Energy Kalman Filter (MEKF) are designed using two different attitude parameterizations i.e. Euler Angle (EA) and Unit Quaternion (UQ). The two filters AE-UKF and UQ-MEKF use highly nonlinear process and measurements estimators, zero-mean Gaussian noises, and first-order Gauss-Markov biases models. Field experiments, adding systems, and data fusion algorithms are used to achieve real-time tracking and estimation. The results show that both AE-UKF and UQ-MEKF demonstrate high accuracy estimation and robustness for UAVs attitude as well as MEMS biases filtering.