Indoor Positioning without Absolute Position Information

M.Sc. Lukas Köping, University of Applied Sciences Würzburg

Date: 25/09/2013
Time: 10:00-11:00
Room: H-F 012

Today's indoor positioning systems for pedestrians mostly make use of absolute positioning with WiFi signals. While such systems can lead to good position estimates, high effort is needed to set them up. In an off-line phase the received signal strengths (RSS) are measured at different locations and listed in a radio map. During the localization process the signal strengths at the current position of the target object are measured and compared against the radio map entries. Thereby every change in the environment forces to repeat the process of setting up a radio map.

A different approach is to use dead reckoning. Given a starting position one can use the data of an accelerometer, a gyroscope and a magnetometer to estimate the travelled distance and the walking direction.  While with this data a new position for the current time step can be calculated, usually the resulting position is highly erroneous due to the imprecision of the sensors. 

In this talk, a method is presented that utilizes step- and turn detection together with floor map information to overcome this sensor imprecision. The fusion of the data takes place using particle filtering. In contrast to state of the art methods, observation data is integrated into the state transition model and the information of the previous time step is used to improve the observation model. Experimental results will be presented that show the accuracy of the proposed method.