Active and Adaptive Dive Planning for Dense Bathymetric Mapping

Abstract

We examine the problem of planning dives for an Autonomous UnderwaterVehicle (AUV) to generate a dense bathymetric map using sidescansonar. The three key challenges in this scenario are (1) proper modelingof the local uncertainty of the 3D reconstruction, (2) efficient diveplanning to reduce this uncertainty, and (3) determination of whento re-plan adaptively based on new information. To address thesechallenges, we propose using non-parametric Bayesian regression tomodel the expected accuracy of the map, which provides principledcost functions for planning subsequent dives. In addition, we proposean efficient greedy method to reduce this uncertainty, and we showthat it achieves theoretically bounded performance given assumptionson the sensor model and the form of the uncertainty function. Wepresent experiments on the propeller-driven YSI EcoMapper AUV equippedwith a sidescan sonar in an inland lake. The experiments demonstratethe benefit of efficient dive planning, with our results providingperformance gains of up to 83% versus standard lawnmower patterns.