Scholarlink

To address the challenges of rapid perception and environmental modeling in unstructured three- dimensional dynamic adversarial environments, including backward technology, low information resolution, and poor real-time data collection, this paper utilizes Unmanned Ground Vehicle (UGV) equipped with 3D LiDAR scanners and low-altitude Unmanned Aerial Vehicle (UAV) equipped with visible light cameras for air-ground integrated rapid perception and mapping of complex environments. The system achieves precise three- dimensional registration, multi-source information fusion, rapid perception, and real-time reconstruction of complex environments and typical targets. For joint modeling, an improved FAST-LIO algorithm is employed to construct LiDAR point cloud maps and visual feature maps for both ground unmanned platforms and UAVs, while ICP registration algorithm fuses point cloud and feature maps to generate high-precision maps with wide range, broad field of view, and rich features. For joint perception, real-time analysis of video images from UGV LiDAR, and UAV cameras enables intelligent target detection and recognition based on YOLO-V3 deep learning algorithm, followed by multi-hypothesis theory and PCR5 fusion rules from DSmT theory to fuse LiDAR and camera recognition results. Experimental validation using the proposed multi-agent air-ground joint mapping and perception method in urban and mountainous terrains demonstrates effectiveness and stability, meeting practical requirements for unstructured three-dimensional dynamic environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on Modeling Methods for Unstructured Three- Dimensional Dynamic Adversarial Environments

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mianhao Qiu,
  • Xingyang Feng,
  • Jing Luo,
  • Hua Cong,
  • Shaoliang Zhang

摘要

To address the challenges of rapid perception and environmental modeling in unstructured three- dimensional dynamic adversarial environments, including backward technology, low information resolution, and poor real-time data collection, this paper utilizes Unmanned Ground Vehicle (UGV) equipped with 3D LiDAR scanners and low-altitude Unmanned Aerial Vehicle (UAV) equipped with visible light cameras for air-ground integrated rapid perception and mapping of complex environments. The system achieves precise three- dimensional registration, multi-source information fusion, rapid perception, and real-time reconstruction of complex environments and typical targets. For joint modeling, an improved FAST-LIO algorithm is employed to construct LiDAR point cloud maps and visual feature maps for both ground unmanned platforms and UAVs, while ICP registration algorithm fuses point cloud and feature maps to generate high-precision maps with wide range, broad field of view, and rich features. For joint perception, real-time analysis of video images from UGV LiDAR, and UAV cameras enables intelligent target detection and recognition based on YOLO-V3 deep learning algorithm, followed by multi-hypothesis theory and PCR5 fusion rules from DSmT theory to fuse LiDAR and camera recognition results. Experimental validation using the proposed multi-agent air-ground joint mapping and perception method in urban and mountainous terrains demonstrates effectiveness and stability, meeting practical requirements for unstructured three-dimensional dynamic environments.