Previous Abstract Return to Session C5 Next Abstract

Session C5: Navigation Using Environmental Features

Enhanced Acceleration Phase Tracking for Moving Platform Detection in 3D Indoor Navigation
Susanna Kaiser, German Aerospace Center (DLR), Germany
Location: Windjammer

Tracking pedestrians by the use of inertial measurement units (IMUs) is widespread due to the fact that low cost inertial sensors can be found in every smartphone or other carry on electronic device. Especially in applications, where only inertial sensors are available for navigation due to missing infrastructure or privacy reasons e.g. not allowing cameras, pedestrian navigation based on only inertial sensors comes into play. Moreover, in indoor environment, urban canyons, tunnels or even mines Global Navigation Satellite Systems (GNSS) might be disturbed or even lost, therefore, pedestrian dead reckoning (PDR) based on inertial sensors helps to overcome this problem.
Unfortunately, with inertial PDR the full 3D localization is currently not satisfactory solved due to the fact that the height estimated by an inertial sensor drifts and it ignores any user displacement different from doing steps, such as by moving platforms. The first problem can be solved by sensor fusion with barometric measurements, but the second problem of ignoring the moving platforms is not yet fully solved. Examples of moving platforms are elevators, escalators or moving sidewalks e.g. in an airport. Due to the fact that the inertial sensor errors accumulate, PDR systems are usually adapted to the human gait pattern in order to constrain the error. This adaptation leads to ignoring the moving platform as it is the case with the so called zero velocity updates at every step in the strapdown algorithm assuming foot mounted inertial sensors. It is also the case when applying a combined orientation estimation and step and heading estimation algorithm commonly used when the sensor is placed at another body location than the foot.
In order to overcome this problem, a moving platform detection (MPD) algorithm was proposed in [1] as a first step to handle moving platforms. In this algorithm mainly three different detection techniques are applied: detection based on the acceleration phase, on the magnetic variance and on the barometric pressure. The acceleration phase tracking module is based on a constraint open-loop integration of the acceleration signals in the navigation frame (i.e. the global coordinate system). Recently, magnetometers and barometers are commonly embedded together with the accelerometers and gyroscopes of the inertial sensor. Therefore, it is very common to use also those signals and they can also be easily utilized for the MPD. The magnetic variance detector exploits the fact that the disturbance sources of the magnetometer measurements are changing during the ride on a moving platform. The barometric pressure is investigated by applying a simple linear regression line and investigating if it fits to the barometric measurements.
In this paper, we look more deeply at the acceleration phase tracking module. The acceleration phase tracking module of the MPD indicates the boundaries of the interval being on a moving platform by investigating the velocity applying a threshold. For the acceleration phase tracking it is very crucial to perform a very accurate technique for estimating the biases of the acceleration signals. Any error in the bias estimation will cause a drift in the estimated velocity signal, which in turn will cause a false detection or a non-detection of the moving platform. In this paper, we will investigate different bias estimation techniques like taking the mean value during the stance phases, applying an auto regressive model or making use of the bias calculation of the Unscented Kalman Filter applied in the strapdown algorithm. We will compare the outputs of the different algorithms and will deeply investigate the influence of the bias estimation on the acceleration phase tracking module. The acceleration phase tracking module will extensively be tested by a series of measurements dealing with different escalators and elevators rides in different buildings.
[1] Kaiser, S. and Lang, C., “Detecting Elevators and Escalators in 3D Pedestrian Indoor Navigation”, 2016 International Conference on Indoor Positioning and Indoor Navigation (IPIN), 4-7 October 2016, Madrid, Spain



Previous Abstract Return to Session C5 Next Abstract