Eventos Anais de eventos
DINAME 2023
XIX International Symposium on Dynamic Problems of Mechanics
Low-cost inertial sensor fusion with the Ensemble Kalman filter for ground vehicle position estimation
Submission Author:
Flávio Trigo , SP
Co-Authors:
Enrico Bertolini Carlini, Agenor Fleury, Flávio Trigo
Presenter: Flávio Trigo
doi://10.26678/ABCM.DINAME2023.DIN2023-0035
Abstract
Usually, the position of ground vehicles is obtained from the Global Navigation Satellite System (GNSS), which is a source of reliable estimates; however it is prone to signal loss specially in dense urban areas, where buildings shadow the direct reception of satellite signals. An alternative to overcome this drawback is dead reckoning: Integration of low-cost encapsulated inertial measurement units' (IMUs) signals (angular velocity, from gyro, and acceleration, from accelerometer) in order to estimate the target position. This method, however, results in error build-up due to model uncertainties and measurement noise. Therefore, the research on an uninterrupted and trustworthy estimation method is relevant to the navigation field. In this work, we implement a non-linear discrete quaternion-based model to describe the kinematics of a rigid body with six degrees of freedom and apply the Ensemble Kalman filter (EnKF) to fuse IMU sensors' signals and perform dead-reckoning. The estimation method is evaluated in a field test with a passenger car traveling on a 370-meter long closed trajectory. Measurements of linear and angular velocities, and barometric pressure, are the observations processed by the EnKF, whose estimates are compared with ones from the Unscented Kalman filter (UKF). The EnKF position estimation discrepancy on the departure/arrival locus was 2 m, whereas the maximum error along the track was 8 m. Those preliminary results suggest that the EnKF might be a viable option to perform dead-reckoning. We currently seek to optimise the EnKF estimates by fusing data from several IMUs.
Keywords
data fusion, dead reckoning, Ensemble Kalman Filter, stochastic parameter estimation, inertial sensors
