Doctor's Theses (authored and supervised):
E. Pulido Herrera:
"Improving Data Fusion in User Positioning Systems";
Supervisor, Reviewer: R. Quiros, G. Fabregat, H. Kaufmann;
Institute for Software Technology & Interactive Systems,
A fault detection and correction methodology for user positioning systems based on Kalman filtering is presented in this thesis. To develop this methodology,various aspects of the design of positioning systems are taken into account, such as the dynamic models of the systems, the definition of the estimation technique, the available technology and the environment (to identify disturbances that can affect the systems). In the first phase, the available technologies were identified, and the algorithms and the physical-mathematical models were defined for the positioning systems for both indoor and outdoor environments. The Dead Reckoning (DR) algorithm was included for this reason, because it can be applied in both environments. This algorithm allows the pedestrianīs position to be obtained while s/he is walking. Given that one of the main parameters of DR is the pedestrianīs step length, a careful analysis was carried out to determine it. Algorithms that allow, first, detection of a step and, second,calculation of its length are therefore presented here. An integrated system based on UWB and inertial technologies (IMU) is proposed for indoor environments. This system uses the information about the step length to improve the information provided by the UWB system. The system that was defined for outdoor environments is a GPS-IMU system based on the DR algorithm. Data fusion is carried out by means of Kalman filtering for both systems. In the GPS-IMU-DR system, the errors of the azimuth bias and the step length are obtained by means of Kalman filtering, which allows the DR parameters to be corrected and, consequently, the pedestrianīs position can be obtained with greater accuracy. In the next phase, the fault detection and correction methodology is developed. This is based on the principles of causal diagnosis using the theory of possibility. This methodology is proposed in order to prevent the introduction of erroneous information into the Kalman filter. In order to carry this out, failure states of the sensor systems are defined and corrective measures are applied when one or more of those states are present. These states are defined taking into account the empirical knowledge of the behaviour of the system. The performance of the filters was also monitored. This consists in the continuous evaluation of their innovations and, in case of inconsistency, corrective measures are applied to the parameters of the filters. The experiments, for the proposed systems, presented results that improved their initial response to a considerable degree. For the UWB-IMU system, an analysis that involves the detection of inconsistencies of the filter is presented, while for the GPS-IMU-DR system the fault detection and correction methodology is applied. In the UWB-IMU system, the corrective measures introduced into to the Kalman filter allowed both consistent filtering and a soft signal to be obtained, i.e. most of the reflections of the signal provided by the UWB system were eliminated. The results of the GPS-IMU-DR system indicated that by implementing the methodology developed here, consistent filtering was obtained and the values of the azimuth and the step length were corrected properly. As a result more exact pedestrian trajectories were obtained.
Data fusion, Sensors, Tracking
Electronic version of the publication:
Created from the Publication Database of the Vienna University of Technology.