Scan Matching for Flooded Subterranean Voids

Abstract

This paper presents a new technique for scan matching with sparse, noisy sensors such as sonar range finders. This method allows an underwater vehicle to estimate its position within a 2-D plane. When combined with other sensors typically used on such vehicles, full 6 degree of freedom pose estimation is possible in flooded subterranean environments. Our approach uses a particle filter and an approximation of the likelihood of sensor readings, based on nearest neighbor distances, to approximate the probability distribution over possible poses. We present simulated results based on field collected data, demonstrating our approaches advantages over traditional scan matching techniques.