Towards Semantic-Aware Active Gas Distribution Mapping in Unknown Cluttered Environments

Mal Fazliu, Cunjia Liu

Keywords: gas distribution mapping, unknown environments, semantic occupancy mapping, informative path planning, exploration-exploitation, semantic-aware planning

TL;DR: A semantic-aware extension to active gas distribution mapping: a real-time 3D semantic occupancy map provides exploration cues (gas-relevant landmarks) and physics cues (permeable structures) that pure geometry cannot.

Abstract: Active gas distribution mapping (GDM) in unknown cluttered environments requires a mobile robot to simultaneously explore its surroundings, build an occupancy map, and infer a gas concentration field from sparse chemical measurements. Existing active GDM frameworks treat the environment as purely geometric, ignoring what the obstacles actually are. We present an ongoing effort to integrate real-time 3D semantic perception into active GDM. A semantic voxel mapping module fuses YOLO instance segmentation on RGB imagery with aligned depth measurements, maintains a per-voxel class belief, and prunes against an online OctoMap, producing a live semantic layer aligned with the occupancy representation. Such a representation can be used in conjunction with our Exploration and Exploitation Informed Trees (XIT) planner for a semantic-aware active GDM pipeline.

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: 7