Autonomous navigation method for agricultural robots in high-bed cultivation environments

• Novel navigation method switches between waypoint and cultivation bed navigation. • Proposed method minimizes reliance on complex planning in structured farming. • Achieved ± 0.05 m accuracy in bed-to-robot distance during autonomous navigation. • Demonstrated robust navigation in virtual and real-world environments. • The importance of testing agricultural robots in virtual environments is discussed. This study proposes an autonomous navigation method for agricultural robots designed for high-bed cultivation. The proposed method integrates two navigation strategies: waypoint navigation, which directs the robot to a predefined waypoint, and cultivation bed navigation, which ensures precise movement between cultivation beds. By alternating between these navigation methods, the robot can achieve self-navigation within a farm without relying on path planning, which requires accurate localization in areas with limited environmental features. The robot uses light detection and ranging (LiDAR) point cloud data to navigate effectively. The navigation approach was initially simulated in a virtual environment and then evaluated in a real-world strawberry farm. The results demonstrated the ability of the robot to maintain a specified distance of ± 0.05 m and an orientation angle of ± 5° relative to the cultivation bed. These findings confirm the feasibility of the proposed method for achieving accurate and stable navigation on a farm. This study also highlights the importance of simulations in agricultural robotics development. Simulated environments provide a cost-effective platform for refining robot specifications, such as sensor selection and navigation algorithms, before real-world deployment. For example, simulations have shown that reducing the maximum measurement range of the LiDAR can significantly impact localization accuracy and navigation stability. Future work will focus on creating dynamic simulation environments that replicate real-world conditions, such as uneven surfaces and varying farm layouts. Enhancing simulation fidelity will improve the reliability of evaluations and accelerate the practical implementation of agricultural robots, contributing to their broader adoption and efficiency in farming operations.