Fine-Tuning Localization and Navigation Parameters for Accurate Greenhouse Mapping

Abstract

Accurate mapping is crucial for maximizing productivity and sustainability in agriculture. However, creating maps in greenhouse environments is challenging due to their intricate layouts, often resulting in bending and reduced precision. To address these challenges, fine-tuning was applied to the parameters of the Simultaneous Localization and Mapping (SLAM) algorithm and the navigation process, specifically focusing on the Grid-based Mapping (GMapping) and Dynamic Window Approach (DWA) techniques. SLAM experiments were conducted in a simulated greenhouse environment created by Gazebo, with all operations executed under the Robot Operating System (ROS) framework, enabling real-time mapping and localization. Comparisons between maps generated with and without fine-tuning, and the Gazebo reference map, show a 77.8% improvement in reducing map distortion, resulting in more precise greenhouse representations. These findings highlight how the fine-tuning of algorithm parameters can improve mapping accuracy, ultimately enhancing agricultural applications. Future work will focus on testing this methodology to ensure broader applicability and reliability in real-world greenhouse environments.