A Probabilistic Framework for Sonar Scan Matching Localization

Mobile Robot Localization using Particle Filters and Sonar Sensors

“…Also, Groβmann et al (Groβmann & Poli 2001) confront the sonar localization problem by means of the Hough transform and probability grids to detect walls and corners. Burguera et al (Burguera et al 2008a) adopt a different approach, named spIC (sonar probabilistic Iterative Correspondence), not requiring environmental features to be detected. They have shown that scan matching localization, even in its most basic expression (Burguera et al 2005), can be applied to sonar sensors if their limitations and uncertainties are appropriately taken into account.…”

“…There exist many algorithms to match sets of range readings in the scan matching literature] (Lu & Milios 1997;Rusinkiewicz & Levoy 2001;Pfister et al 2004;Burguera et al 2008a). Most of them follow the structure proposed by the ICP algorithm.…”

“…This algorithm, which has been vastly used by the localization community, has also proved to be effective when applied to sonar readings (Burguera et al 2005). However, recent works show that other matching approaches are able to provide more accurate and robust estimates (Burguera et al 2008a;Burguera et al 2008b). In consequence, it is reasonable to assume that the presented particle filter approach could benefit of these recent matching techniques in the measurement model.…”

Pose-based SLAM with probabilistic scan matching algorithm using a mechanical scanned imaging sonar

“…Both situations might prevent a possible association or even generate a wrong one. The spIC algorithm proposed in [11] is a statistical extension of the ICP algorithm where the relative displacement q as well as the observed points in both scans r i and n i , are modeled as random Gaussian variables (r.g.v.). For a better understanding the algorithm is reproduced in Algorithm 1.…”

“…The probabilistic Iterative Correspondence method (pIC), proposed in [10], explicitly deals with those uncertainties to decide which points in the reference scan are statistically compatible with a certain point of the new scan. A probabilistic weight average is used then to compute a virtual association point for the matching during the minimization [11]. Although the method is suitable for laser data, the same authors have noted that for sparse noisy data like sonar, better results can be achieved using the ICNN data association algorithm instead of using the virtual point (spIC).…”