Discrete-Time Markovian Jump Linear Robust Filtering for Visual and GPS Aided Magneto-Inertial Navigation

Sensor fusion is a major field in autonomous mobile robots navigation research. These methods integrate information obtained from accelerometers, rate gyros and monocular cameras to provide pose and orientation of the robot, which are known in the literature as Visual-Inertial Simultaneous Localization and Mapping systems. For outdoor navigation, sensor fusion algorithms may also use magnetometers and GPS modules, since in indoor environments and certain urban areas they may suffer from measurements corruption in the presence of ferromagnetic materials and signal occlusion, respectively. To avoid combining corrupted or noisy data, we propose a Robust Kalman Filter (RKF) for Discrete-time Markovian Jump Linear Systems which estimates the position and orientation of the platform considering the best Markovian mode at each instant. The proposed RKF approach reduces the degradation of the filter performance due to parametric uncertainties present in the system models. We present experimental results in comparison with standard and state-of-the-art sensor fusion techniques to show that our system is robust even in challenging conditions.