Memory-Efficient Boundary Map for Large-Scale Aerial Maritime Inspection

Benxu Tang, Yunfan Ren, Yixi Cai, Fanze Kong, Wenyi Liu, Fangcheng Zhu, Longji Yin, Liuyu Shi, Fu Zhang

Keywords: Occupancy Grid Mapping, Large-Scale Autonomous Navigation

TL;DR: Our "boundary map" represents environments using 2D boundary voxels rather than 3D volumes , drastically reducing memory overhead to enable large-scale autonomous maritime inspection on aerial robots.

Abstract: Performing tasks in maritime infrastructures poses significant safety risks to human operators. Autonomous aerial robots offer a safer alternative. To achieve such autonomy, an efficient and robust representation of the environment is fundamental. Occupancy grid maps provide an effective solution; however, in large-scale environments, existing methods impose high memory and computational demands, rendering them unsuitable for deployment on resource-constrained aerial platforms. To tackle this, we utilize a novel 2D boundary voxel representation instead of explicitly representing entire 3D volumes. We term our method as boundary map. This low-dimensional representation significantly reduces memory consumption by orders of magnitude, as demonstrated in our benchmark experiments. Furthermore, we validate our framework through long-range autonomous navigation experiments in a GNSS-denied, multi-floor real-world environment. Our results demonstrate that the proposed efficient mapping framework has high potential for supporting complex, large-scale maritime inspection tasks where onboard memory and computational resources are limited.

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: 19