Topology-Based Evaluation for Symbolic Indoor Positioning Algorithms

Abstract

Smart factories require location aware services such as asset tracking. These location aware services should be based on indoor positioning systems. Symbolic indoor positioning is considered as a classification task, where each category denotes a well-defined part of the building, such as a room or corridor. Hence, standard classifiers can be applied to symbolic indoor positioning. A topology-based classification evaluation method is presented that calculates the classification error based on the gravitational force between the symbolic positions denoted by categories. Three variants of the proposed topology-based method is evaluated and compared to the CRISP approach. The comparison was performed over a dataset recorded in a three-storey building whose topology is given in Indoor Geographic Markup Language format. Experimental results showed that the topology-based method gives a more detailed comparison of classifiers for indoor positioning than CRISP.