Estimation of attitudes from a low-cost miniaturized inertial platform using Kalman filter-based sensor fusion algorithm

Abstract

Due to costs, size and mass, commercially available inertial navigation systems are not suitable for small, autonomous flying vehicles like ALEX and other UAVs. In contrast, by using modern MEMS (or of similar class) sensors, hardware costs, size and mass can be reduced substantially. However, low-cost sensors often suffer from inaccuracy and are influenced greatly by temperature variation. In this work, such inaccuracies and dependence on temperature variations have been studied, modelled and compensated in order to reach an adequate quality of measurements for the estimation of attitudes. This has been done applying a Kalman Filter-based sensor fusion algorithm that combines sensor models, error parameters and estimation scheme. Attitude estimation from low-cost sensors is first realized in a Matlab/Simulink platform and then implemented on hardware by programming the micro controller and validated. The accuracies of the estimated roll and pitch attitudes are well within the stipulated accuracy level of ±5° for the ALEX. However, the estimation of heading, which is mainly derived from the magnetometer readings, seems to be influenced greatly by the variation in local magnetic field.