Comparison of Interval Constraint Propagation Algorithms for Vehicle Localization.docVIP

A Journal of Software Engineering and Applications, 2012, 5, 157-162 doi:10.4236/jsea.2012.512b030 Published Online December 2012 (/journal/jsea) Comparison of Interval Constraint Propagation Algorithms for Vehicle Localization I. K. Kueviakoe , A. Lambert , P. Tarroux 1,2 3 1,4 1 LIMSI, CNRS, 91400 Orsay, France; 2 Univ. Paris Sud, IEF, 91400 Orsay, France; 3 LIVIC, IFSTTAR, 14 route de la minière, 78000 Versailles,France; ENS, 45 rue dULM, 75230 Paris, France 4 Email: kangni.kueviakoe@limsi.fr Received 2012 ABSTRACT Interval constraint propagation (ICP) algorithms allow to solve problems described as constraint satisfaction problems (CSP). ICP has been successfully applied to vehicle localization in the last few years. Once the localization problem has been stated, a large class of ICP solvers can be used. This paper compares a few ICP algorithms, using the same expe- rimental data, in order to rank their performances in terms of accuracy and computing time. Keywords: Interval Analysis; Constraint Propagation; Data Fusion; Vehicle Positioning, GPS. 1. Introduction AC-4 to non binary constraints. In 1987, Cleary [12] and Davis [13] were the first to deal jointly with constraint propagation on interval analysis. In 1989, Hyvonen [14] designed a generalized constraint propagation scheme based on interval arithmetic. A narrowing algorithm was proposed by Cleary [12] and improved by Benhamou [15]. Later, that algorithm was renamed HC3 [16] be- cause it is very close to the classical AC-3. Next, Ben- hamou proposed HC4 [16] that does not need the de- composition of constraints. However HC4 suffers from the dependency problem (see section 2.3). To cope with this problem, BC3 [17] was developed but suffers of slowness. BC4 [16] merges both BC3 and HC4 and re- duces the computation time. BC5 [18], by using the in- terval Gauss-Seidel m