[Back]

M. Thienelt, A. Eichhorn, A. Reiterer:
"Intelligent Pedestrian Positioning in Vienna: Knowledge-based Kalman Filtering";
Talk: 5th International Symposium on Mobile Mapping Technology 2007, Padua, Italy; 2007-05-29 - 2007-05-31; in: "Proceedings of the 5th International Symposium on Mobile Mapping Technology (CD); FIG/ISPRS University of Padua, Interdepartment Research Center for Geomatics (Hrg.)", (2007), 7 pages.

In the field of vehicle navigation the combination of position measurements with digital maps usually increases the accuracy and availability of position information. Usually the combination is realised by map-matching techniques, this means fitting the vehicles trajectory into the road network by curve matching. However pedestrians frequently move outside of digitally acquired road networks. Consequently a transfer of the above procedure is not always possible. Typical "off-road" scenarios are represented by pedestrian precincts, parkways and by the whole indoor area. In contrast to vehicle positioning fixing the pedestrian's trajectory to specific ways is not possible due to his more variable possibilities to move. Nevertheless the increasing importance of Location Based Services (LBS) requires an accurate, reliable and preferably always available determination of the pedestrianīs position information. In passive environments usually absolute and relative position sensors are available which are integrated into the userīs mobile device (e.g. PDA). In most cases the sensors are GPS, altimeter, step counter and digital compass. Digital maps are used for the visualization of the position based on the sensor outputs. In this paper the prototype of a map-independent knowledge-based Kalman filter ("WiKaF") for optimal pedestrian positioning is presented. The WiKaF concept, its system architecture and the integrated sensors are described. The multi-sensor system comes from the NAVIO project (another project for pedestrian navigation in Vienna) and contains a Dead Reckoning Module DRM III, a barometer PTB 220, a digital compass HMR 3000 and an eTrex Summit GPS receiver. The two main components of the positioning module are introduced. The knowledge-based component is responsible for the pre-filtering process of the measurement data which includes a first step of outlier detection. It also manages one part of the sensor calibration (determination of scale factors, offsets, etc.). In a further step the central Kalman filter derives the optimal position of the pedestrian. For support in dead reckoning scenarios the filters system and measurement equations connect the multi-sensor output with a causative motion model. The combination of knowledge-based component and Kalman filtering aims at an increasing reliability and an additional reduction of the inertness of the filter. At the end of the paper some promising outdoor and indoor test results of the developed positioning module are presented.