Friedrich M. Wahl

Efficient Parallel Random Sample Matching for Pose Estimation, Localization, and Related Problems

This paper introduces a parallel variant of the Random Sample Matching (RANSAM) approach leading to a very time and memory efficient enhancement of the traditional Random Sample Consensus (RANSAC). This improvement exploits the theory of the birthday attack whose mathematical background is known from cryptography. The RANSAM technique can be applied to a wide field of applications, for example, localization in mobile robotics and object pose estimation in automated assembly. Standard computers feature multi-core processors nowadays - hence a considerable speedup can be obtained by distributing selected subtasks of RANSAM among the available processors. Primarily, this paper addresses the parallelization of the RANSAM approach and the selection of the quality function that evaluates the validity of a matching - since each application scenario and sensor setup requires an individually appropriate quality measure. Apart from general performance results, experimental results regarding a special application - pose estimation with tactile sensors - are presented.