Abstract
High-resolution soil-health assessment requires repeatable measurements across agricultural fields. However, manual soil sampling is slow, labor-intensive, and inconsistent, and limited sampling coverage can miss spatial and depth variability. This study developed and field-tested an RTK GNSS-guided waypoint navigation system for a Clearpath Husky A300 unmanned ground vehicle. The system is intended to support autonomous soil sampling and future in-field soil sensing.
The navigation framework was implemented in ROS 2. At startup, the system loaded predefined GPS targets and waited for valid GPS data and stable inertial navigation system heading. The rover used inertial heading feedback for point-turn alignment before each path segment. During tracking, a proportional-derivative controller used GPS-based movement heading, cross-track error, and heading error to generate steering and speed commands. The rover stopped for five seconds at each waypoint and recorded navigation data. A progressive forward and reverse rocking strategy helped the rover recover when it could not complete point turns on loose soil.
Field trials evaluated grid, circular, and irregular waypoint layouts. Mean absolute cross-track error was 5.88 cm for the grid path, 19.46 cm for the circular path, and 7.88 cm for the irregular path. Average waypoint stopping error was 9.45 cm, 9.42 cm, and 9.48 cm, respectively. The grid run produced the lowest mean absolute cross-track error among the displayed trials, although these individual runs do not establish statistical differences.
The results demonstrate the feasibility of combining precise waypoint navigation, adaptive turning recovery, and soil-sensing development for autonomous agricultural field operations. Future work will include repeated navigation trials, obstacle avoidance, route optimization, sensor integration, and real-time soil-data visualization.
Keywords
RTK GNSS, autonomous navigation, unmanned ground vehicle, agricultural robotics, soil sampling, waypoint navigation, adaptive turning, precision agriculture
DOI
10.5703/1288284318717
RTK GNSS-Guided Waypoint Navigation with Adaptive Turning for Autonomous Soil Sampling
High-resolution soil-health assessment requires repeatable measurements across agricultural fields. However, manual soil sampling is slow, labor-intensive, and inconsistent, and limited sampling coverage can miss spatial and depth variability. This study developed and field-tested an RTK GNSS-guided waypoint navigation system for a Clearpath Husky A300 unmanned ground vehicle. The system is intended to support autonomous soil sampling and future in-field soil sensing.
The navigation framework was implemented in ROS 2. At startup, the system loaded predefined GPS targets and waited for valid GPS data and stable inertial navigation system heading. The rover used inertial heading feedback for point-turn alignment before each path segment. During tracking, a proportional-derivative controller used GPS-based movement heading, cross-track error, and heading error to generate steering and speed commands. The rover stopped for five seconds at each waypoint and recorded navigation data. A progressive forward and reverse rocking strategy helped the rover recover when it could not complete point turns on loose soil.
Field trials evaluated grid, circular, and irregular waypoint layouts. Mean absolute cross-track error was 5.88 cm for the grid path, 19.46 cm for the circular path, and 7.88 cm for the irregular path. Average waypoint stopping error was 9.45 cm, 9.42 cm, and 9.48 cm, respectively. The grid run produced the lowest mean absolute cross-track error among the displayed trials, although these individual runs do not establish statistical differences.
The results demonstrate the feasibility of combining precise waypoint navigation, adaptive turning recovery, and soil-sensing development for autonomous agricultural field operations. Future work will include repeated navigation trials, obstacle avoidance, route optimization, sensor integration, and real-time soil-data visualization.