2019/02/25 by Marija Popović, Popovic, Marija, Teresa Vidal‐Calleja +7 · 5 citations
Computer Science · Environmental Science · #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Robotics (cs.RO) #Species Distribution and Climate Change
paper · pdf · doi:10.48550/arxiv.1902.09660
openalex publication_date 2019/02/25 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
Information gathering algorithms play a key role in unlocking the potential\nof robots for efficient data collection in a wide range of applications.\nHowever, most existing strategies neglect the fundamental problem of the robot\npose uncertainty, which is an implicit requirement for creating robust,\nhigh-quality maps. To address this issue, we introduce an informative planning\nframework for active mapping that explicitly accounts for the pose uncertainty\nin both the mapping and planning tasks. Our strategy exploits a Gaussian\nProcess (GP) model to capture a target environmental field given the\nuncertainty on its inputs. For planning, we formulate a new utility function\nthat couples the localization and field mapping objectives in GP-based mapping\nscenarios in a principled way, without relying on any manually tuned\nparameters. Extensive simulations show that our approach outperforms existing\nstrategies, with reductions in mean pose uncertainty and map error. We also\npresent a proof of concept in an indoor temperature mapping scenario.\n