Small UAV localization based strong tracking filters augmented with interacting multiple model

Abstract

The most research work of small unmanned aerial vehicles (SUAVs) localization problem relies on the fusion of data from multiple on board sensors by utilizing different Kalman filtering techniques based on a single complex model. Here, the interacting multiple model technique (IMM) augmented by four simple kinematic models has been proposed to provide good estimation performance. The proposed IMM incorporates the models of constant velocity (CV), constant acceleration (CA), horizontally coordinated turn (HCT) and 3D coordinated turn (3DCT). Each model represents a specific motion part for the flight trajectory and attached by such an adaptive robust filter. The constant velocity and acceleration models are attached with strong tracking Kalman filter (STKF) while, the horizontally coordinated turn and 3D coordinated turn models are attached with Strong tracking Extended Kalman filter (STEKF) because they are nonlinear and the turn rates are not constant. Strong tracking filtering process based on an interacting multiple models algorithm (STF-IMM) has been proposed to estimate the position and velocity information for small UAV to improve the accuracy of the estimation algorithm performance and satisfy the robustness. The state estimation is performed by switching between multiple models based on Markov process. The proposed technique provides an advantage when the dynamic behavior of the system cannot be characterized by a single model and also improves the system robustness against modeling defects. The proposed STF-IMM approach is implemented and compared with a single complex model based on two different navigation filtering. The proposed algorithm has been tested in sophisticated trajectory. The simulation results and comparison clearly show that the proposed STF-IMM approach performs better than the other navigation filtering based single complex model and provides a more accurate estimation of position and velocity. Also, the simulation results proved that the proposed approach is well suited to yield an improvement in the estimation accuracy of small UAV localization.