Considerations on quality metrics for self-localization algorithms

  • Authors:

  • Juergen Eckert;Felix Villanueva;Reinhard German;Falko Dressler

  • Affiliations:

  • Computer Networks and Communication Systems, Dept. of Computer Science, University of Erlangen, Erlangen, Germany;Computer Architecture and Networks, School of Computer Science, University of Castilla-La Mancha, Ciudad Real, Spain;Computer Networks and Communication Systems, Dept. of Computer Science, University of Erlangen, Erlangen, Germany;Computer Networks and Communication Systems, Dept. of Computer Science, University of Erlangen, Erlangen, Germany

  • Venue:
  • IWSOS'11 Proceedings of the 5th international conference on Self-organizing systems
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

The demand for location awareness and, therefore, the demand for self-localization techniques is continuously increasing. As a result, a good number of systems and methods for self-localization have been developed. Almost every system described in the literature exploits specific hardware or scenario features to solve the positioning issue, e.g. by using anchor nodes, relying on distances or angles, and even focusing on quite different distances ranging from centimeters to several kilometers. In many cases, the metrics used to evaluate the localization quality have been chosen according to the scenario. In this paper, we thoroughly discuss the most frequently used metrics for evaluating the quality of self-localization techniques. According to our findings, careful handling of some commonly used metrics is strongly required. We further propose an area-based solution that is especially helpful to measure and to compare different localization systems, which only need exact localization in a local context independently from specific scenario or hardware requirements. In this paper, we try to shed light on the question how to compare those very different techniques. In particular, we suggest the use one of two attribute-independent metrics. The first one is a generalization of an already quite popular metric, the Global Energy Ratio (GER), and the latter, the Area Ratio (AR), is a new approach based on the covered area.