In context
By 2025, UAVs were increasingly deployed in large-scale agricultural and forestry tasks, where environmental perception must handle smoke, dust, and variable terrain. Single-sensor mapping—camera or LiDAR—proved fragile, while existing fusion methods either merged point clouds (losing sensor models) or built separate maps per sensor (costly in memory). This paper addressed the need for efficient, robust mapping on embedded UAV computers.
What was reported
Researchers from Shenyang University of Technology and Zhejiang A&F University proposed a direct gridding method to fuse heterogeneous distance sensors—LiDAR and 4D imaging radar—into a single occupancy grid map. The approach uses a ring buffer for sliding map updates, reducing grid count and memory usage. It derives an update error model linking grid resolution, range scaling, and flight speed to ensure safe high-speed operation.
For imaging radar, the method inversely computes scale transformation factors from grid resolution and the radar detection model, allowing direct mapping and alignment of point clouds in a unified grid space without separate calibration. A “probability topographic map” update strategy weights occupancy probabilities with an update coefficient to improve noise resistance in complex environments.
Tests on the NTU4DRadLM dataset under smoke interference showed lower latency and stronger anti-interference than existing methods. In real large-scale terrain missions, memory usage stayed stable at 1.5–2 GB, meeting embedded computing constraints. The sliding update also cut construction time variance by 16 ms.
Why it mattered
This work demonstrated a practical path to robust, low-latency mapping for UAVs in challenging outdoor conditions, balancing sensor fusion benefits with the tight resource limits of onboard computers—an important step for autonomous operations in agriculture, inspection, and exploration.
“The direct gridding method reduces the latency and enhances the resistance to interference while constructing a multi-source heterogeneous fusion-based grid map.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2025-09-16 · “面向无人机的多源异构传感器融合地图建立方法”