Lightweight Smart Sense-and-Avoid Module for Low-Altitude Agricultural UAVs

This paper introduces a lightweight framework aimed at enhancing UAV operations for low-altitude flight in agriculture fields. Farmlands are often filled with obstacles such as trees, structures, and poles, that pose risks of collision during low flying. To address this challenge, the proposed system combines depth data from RGB-D camera with an IMU to generate LiDAR-like data, ensuring seamless integration with any range sensor-based avoidance algorithms. The entire perception and processing stack runs on the sense-and-avoid module, which can be easily mounted on any MAVLink compatible flight controller. This enables the UAV to detect and avoid obstacles in real time, smoothly transitioning from semi-autonomous to fully autonomous flight mode. By minimizing yaw during navigation, the system ensures flight efficiency and stability, using only 40 % of the onboard processor capacity, leaving room for the integration of additional functionalities, such as aerial spraying, data collection, etc. The effectiveness of the system was validated through simulations and real-world tests, showing reliable performance in diverse agricultural settings. This sense-and-avoid technology has broad applications in agriculture, surveillance, and disaster relief, improving UAV navigation in obstacle-dense environments.