In this paper a feasibility study on magnetometer-free sensor fusion in 2D pedestrian position and orientation tracking is presented. The sensor-fusion is performed through a nonlinear Kalman filter variant. The estimated orientation is then used to properly refer measured acceleration in an arbitrarily chosen navigation reference frame. Also, the same data is used to accurately remove the gravitational component from the measurement acceleration before any integration procedure to yield velocity and position. n addition, a custom curve for accurate foot-stance recognition is proposed. A velocity and acceleration reset is performed on the instants recognized thanks to the proposed time series. Estimated position of the foot-mounted IMU is then compared with fixed landmarks. Three walking tasks have been performed: a circle, a square and a longer indoor walk. Results show good accuracy in spite of the absence of heading information coming from the magnetometer (errors below 10 cm for the short trials), although on the long distances drift is more evident (2 m over 48 m of overall walking distance covered). Finally, a series of random walks have been performed to evaluate the foot stance recognition using the proposed time series against the simple accelerometer module evaluation.
Magnetometer-Free Sensor Fusion Applied to Pedestrian Tracking: A Feasibility Study / Cardarelli, S.; Di Florio, P.; Mengarelli, A.; Tigrini, A.; Fioretti, S.; Verdini, F.. - STAMPA. - (2019), pp. 238-242. (Intervento presentato al convegno 23rd IEEE International Symposium on Consumer Technologies, ISCT 2019 tenutosi a ita nel 2019) [10.1109/ISCE.2019.8901014].
Magnetometer-Free Sensor Fusion Applied to Pedestrian Tracking: A Feasibility Study
Cardarelli S.
Methodology
;Mengarelli A.Conceptualization
;Tigrini A.Conceptualization
;Fioretti S.Supervision
;Verdini F.Supervision
2019-01-01
Abstract
In this paper a feasibility study on magnetometer-free sensor fusion in 2D pedestrian position and orientation tracking is presented. The sensor-fusion is performed through a nonlinear Kalman filter variant. The estimated orientation is then used to properly refer measured acceleration in an arbitrarily chosen navigation reference frame. Also, the same data is used to accurately remove the gravitational component from the measurement acceleration before any integration procedure to yield velocity and position. n addition, a custom curve for accurate foot-stance recognition is proposed. A velocity and acceleration reset is performed on the instants recognized thanks to the proposed time series. Estimated position of the foot-mounted IMU is then compared with fixed landmarks. Three walking tasks have been performed: a circle, a square and a longer indoor walk. Results show good accuracy in spite of the absence of heading information coming from the magnetometer (errors below 10 cm for the short trials), although on the long distances drift is more evident (2 m over 48 m of overall walking distance covered). Finally, a series of random walks have been performed to evaluate the foot stance recognition using the proposed time series against the simple accelerometer module evaluation.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.