Research Progress on Multi-Sensor Fusion-Driven Autonomous Navigation and Obstacle Avoidance Technologies for Agricultural Machinery

Autonomous navigation and obstacle avoidance are key technical foundations for intelligent agricultural machinery operating in open fields, orchards, greenhouses, and mixed human-machine production environments. Single-sensor navigation is constrained by canopy occlusion, dust, vibration, illumination variation, reflective soil surfaces, GNSS multipath, terrain unevenness, and the coexistence of crops, workers, vehicles, animals, and implements. Multi-sensor fusion improves positioning continuity, environmental perception, mapping reliability, and real-time safety decisions by integrating GNSS, inertial measurement units, wheel odometry, LiDAR, RGB cameras, RGB-D cameras, stereo vision, radar, ultrasonic sensors, and crop-row perception. Recent studies have advanced from isolated perception modules toward tightly coupled localization, semantic mapping, trajectory tracking, and dynamic obstacle avoidance. Fusion strategies are becoming more adaptive and uncertainty-aware. However, field deployment still faces challenges in environmental robustness, sensor reliability, real-time computation, and perception–control integration. The review summarizes research progress in sensing foundations, calibration, synchronization, fusion architectures, autonomous navigation, obstacle detection, risk prediction, and local planning. The discussion emphasizes sensor characteristics, algorithmic robustness, agricultural scene complexity, and field-deployable autonomy.