4.Conclusions

In this paper, a highly robust autonomous navigation system is proposed for dynamic greenhouse agricultural environments. Initially, a hardware system incorporating multiple sensors is designed and implemented tailored for the greenhouse environment. Then, in order to enhance environmental perception and reduce complex environmental disturbances, a relocalization module is devised and integrated into the SLAM system. Lastly, the local planner of OpenPlanner is enhanced, taking into account additional metrics and introducing a hysteresis strategy for improved performance.The proposed navigation system has been tested in various complex greenhouse environments, demonstrating robust performance throughout all evaluations. As it significantly reduces the absolute pose error in localization, along with ensuring safer trajectories in the path planning process. Consequently, this system exhibits exceptional performance and holds promising potential in the context of greenhouse agricultural scenarios. However, the inclusion of more modules also leads to increased computational demand. Future work may entail the design of additional sensors within the robot hardware system to further improve performance. Moreover, reducing the computational load of the path planning algorithm to enable deployment on more lightweight computational units will be a key focus moving forward.