Scholarlink

<p>This paper proposes an autonomous navigation framework for orchard weeding robots based on the ZTR (Zero Turn Radius) platform. Unlike conventional agricultural robots, our system utilizes multiple sensors, including GNSS, IMU, 2D LiDAR, and cameras, to achieve reliable object detection and accurate position estimation based on the Extended Kalman Filter (EKF). For path tracking, we implemented a pure-pursuit algorithm with a look-ahead distance strategy, which is particularly effective for ensuring stable navigation at low speeds. To evaluate the path tracking performance, experiments were conducted in a structured orchard environment. During autonomous navigation for an average of 8.54&#xa0;min, the proposed framework achieved an average trajectory error of 11.66 cm. For localization performance evaluation, the proposed algorithm was compared with a baseline method that fuses raw GPS and IMU data. The proposed method achieved a mean error of 0.0078 m, representing a 78.3% improvement in accuracy over the baseline.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Field-Validated Navigation Framework for Orchard Weeding Robots Using Multimodal Sensor Fusion

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月28日

查看全文
  • Sangyoon Oh,
  • Hyun-Yong Jeon,
  • Jaehong Seo,
  • Keun Ha Choi,
  • Kyung-Soo Kim

摘要

This paper proposes an autonomous navigation framework for orchard weeding robots based on the ZTR (Zero Turn Radius) platform. Unlike conventional agricultural robots, our system utilizes multiple sensors, including GNSS, IMU, 2D LiDAR, and cameras, to achieve reliable object detection and accurate position estimation based on the Extended Kalman Filter (EKF). For path tracking, we implemented a pure-pursuit algorithm with a look-ahead distance strategy, which is particularly effective for ensuring stable navigation at low speeds. To evaluate the path tracking performance, experiments were conducted in a structured orchard environment. During autonomous navigation for an average of 8.54 min, the proposed framework achieved an average trajectory error of 11.66 cm. For localization performance evaluation, the proposed algorithm was compared with a baseline method that fuses raw GPS and IMU data. The proposed method achieved a mean error of 0.0078 m, representing a 78.3% improvement in accuracy over the baseline.