一种可适配多任务的高效三维地图系统，支持多智能体协同导航。
A Robust, Task-Agnostic and Fully-Scalable Voxel Mapping System for Large Scale Environments
- 基于哈希表的体素结构，按时空优先级管理地图数据
- 实测实现95%以上带宽压缩，支持高分辨率实时建图
- 适合无人机、多机协同等大规模环境导航场景
自主导航在未知环境中的感知仍具挑战性，尤其对空中飞行器而言。现有大多数映射算法针对特定任务设计，限制了扩展与协作应用。本文提出一种体素映射系统，可为多种任务构建自适应地图。系统采用基于哈希表的地图结构，以空间和时间优先级管理每个体素，无需显式地图边界。同时引入高效地图共享机制，带宽消耗极低，支持多智能体应用。我们在真实世界与仿真环境中测试了该系统，应用于局部建图、全局建图、多智能体协同导航及高速导航等多种任务。结果表明，该系统能构建高分辨率、广覆盖、实时响应的可定制地图，无论传感器或环境如何。借助地图共享功能，系统可在保持全分辨率的同时，使传输带宽减少超过95%。
原文摘要 · Abstract (English)
Perception still remains a challenging problem for autonomous navigation in unknown environment, especially for aerial vehicles. Most mapping algorithms for autonomous navigation are specifically designed for their very intended task, which hinders extended usage or cooperative task. In this paper, we propose a voxel mapping system that can build an adaptable map for multiple tasks. The system employs hash table-based map structure and manages each voxel with spatial and temporal priorities without explicit map boundary. We also introduce an efficient map-sharing feature with minimal bandwidth to enable multi-agent applications. We tested the system in real world and simulation environment by applying it for various tasks including local mapping, global mapping, cooperative multi-agent navigation, and high-speed navigation. Our system proved its capability to build customizable map with high resolution, wide coverage, and real-time performance regardless of sensor and environment. The system can build a full-resolution map using the map-sharing feature, with over 95 % of bandwidth reduction from raw sensor data.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。