Stochastic mapping using forward look sonar

The goal of concurrent mapping and localization (CML) is enable a mobile robot to build a map of an unknown environment, while concurrently using that map to navigate. This paper describes the application of the stochastic mapping approach to CML [10, 8, 3] to data from a high resolution array forward look imaging sonar [9, 7, 2]. The objective is to test the feasibility of stochastic mapping as a method for autonomous underwater vehicle (AUV) navigation. Measurements from the forward look sonar were post-processed to yield estimates for the locations of envi-ronmental features and the trajectory of the sensor. The resulting trajectory provides an improvement in comparison to position estimates computed with an inertial navigation system. DGPS data is used as a ground-truth for comparison. The results demonstrate the potential of CML algorithms to achieve a bound-ed navigation error through tracking of environmental features.