Multi-sensor based indoor vehicle and pedestrian navigation

Among the positioning techniques in indoor environments, the approach on the basis of exploiting Wi-Fi is attractive, which is expected to yield a cost-effective and easily accessible solution. Most Wi-Fi localization methodologies rely on the received signal strength (RSS) measurements. In this work, different Wi-Fi RSS based positioning algorithms are explored. The performance of each approach is shown with experimental results.
Considering the complementary nature of Wi-Fi positioning and inertial navigation system (INS), the combination of both systems yields a synergetic effect resulting in higher performance. For indoor vehicle navigation, the performance of the INS/Wi-Fi integrated system can be further improved without hardware change. An enhanced integration, which employs adaptive Kalman filtering (AKF) and vehicle constraints, is presented. The experimental results show that the enhanced integrated system provides higher navigation accuracy, compared to using Wi-Fi positioning and conventional INS/Wi-Fi integration.
For personal navigation applications, the pedestrian dead reckoning (PDR) system is employed. With a foot mounted IMU, zero velocity update (ZUPT) and zero angular rate update (ZARU) methodologies can be applied to re-calibrate the IMU, which can reduce the INS drift errors. For personal navigation with the IMU embedded in the portable device, the adapted PDR based on device placement mode classification is presented. Three typical placement modes are discussed. The classification performances with different classifiers are shown with real test results. The adapted PDR is further combined with Wi-Fi positioning. The experimental results show that the integrated system outperforms the standalone navigation systems.
Attitude estimation is a challenging topic for indoor navigation. The camera based visual gyroscope technique can transform information found from images into the camera rotation. Unlike the rate gyroscope in an IMU, the visual-gyro using vanishing points does not suffer from drift errors. In this work, an INS/visual-gyro integration using direction cosine matrix (DCM) based models is presented. Compared to the conventional Euler angle models, the usage of DCM can provide linear system models and avoid singularity problems. The performance of attitude and gyro bias estimation using the integrated system is shown with turntable test and experimental results.