A comprehensive review of agricultural ground automatic navigation systems based on Multi-sensor fusion

In a complex and changing agricultural environment, achieving autonomous navigation of agricultural robots and field machinery is a fundamental prerequisite for completing various farming tasks. However, because of the unstructured nature, high dynamics, and signal interference commonly found in agricultural scenarios, achieving high-precision autonomous navigation in such challenging environments remains a significant research challenge. In response to the unique complexity of agricultural settings, this study outlines the evolution of agricultural navigation technologies from single-sensor systems to advanced multi-sensor approaches and systematically evaluates the applicability of commonly used single-sensor methods in complex field environments. The findings indicate that, among single-sensor navigation applications, vision-based techniques are more widely adopted in agricultural scenarios. In contrast, for multi-sensor navigation, integrated systems combining inertial navigation systems and global navigation satellite systems are the most prevalent. Furthermore, in terms of multi-sensor fusion algorithms, kalman filters and their extended variants are extensively employed, demonstrating strong robustness and adaptability. These analyses can offer valuable theoretical foundations and practical guidance for the selection and optimization of fusion strategies in future agricultural navigation systems. This study also compared global and local path planning algorithms. Finally, practical applications of fusion technologies in various agricultural scenarios are discussed.