Variable-Resolution Virtual Map Guided Informed Tree Search for Autonomous USV Exploration

Ye Li, Mal Fazliu, Yewei Huang, Alberto Quattrini Li, Brendan Englot, Cunjia Liu, Yuanchang Liu

Keywords: Autonomous Exploration, Active SLAM, Informative Path Planning, Uncertainty-Aware Planning

TL;DR: This paper proposes VRVM-XIT, an uncertainty-guided exploration framework that uses variable-resolution mapping uncertainty to guide informed tree search for GNSS-degraded unmanned surface vehicle exploration.

Abstract:

Autonomous exploration by unmanned surface vehicles (USVs) in GNSS-degraded near-shore environments is challenging. Planning must remain reliable under uneven environmental structure and localization uncertainty. The variable-resolution virtual map (VRVM) provides a scalable way to represent mapping uncertainty. However, when it is combined with a random-sampling RRT planner, the generated trajectories can be highly dispersed and may not follow the uncertainty structure encoded in the map. In this paper, we convert VRVM uncertainty into a continuous information field and use this field to guide tree expansion and trajectory proposal generation.
Our proposed method guides sampling and edge expansion toward informative regions during proposal generation, and selects the best proposal using the virtual-map-based criterion to balance exploration and revisitation. Benchmarking in the VRX Gazebo simulator against VRVM with an RRT-based random-sampling planner shows that the proposed method yields more consistent exploration behavior, lower localization error, higher map coverage, and lower mapping error across the evaluated scenarios.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 12