Sensor-fused EKF-based navigation and kinematic modelling of a remotely controlled mobility vehicle in GPS denied agricultural environments

Accurate row-following is essential for precision agricultural operations such as fertilizer application, inter-row cultivation, and site-specific crop management. However, reliable navigation in agricultural fields remains challenging under GPS-denied conditions caused by dense crop canopy, uneven terrain, and environmental disturbances. This study presents the development and validation of a low-cost sensor-fusion navigation framework for a remotely controlled agricultural mobility vehicle operating in GPS-denied environments. The system integrates wheel encoders and an inertial measurement unit within an extended kalman filter (EKF) for vehicle-state estimation, while ultrasonic sensing is used independently for obstacle detection and occupancy-grid mapping. A differential-drive kinematic model was employed for vehicle motion estimation, while Dijkstra-based path planning and closed-loop trajectory tracking enabled navigation during straight-row traversal and headland turning. Simulation studies in Gazebo and SolidWorks validated the kinematic behaviour and navigation performance before field implementation. Field experiments demonstrated that odometry-only navigation resulted in lateral-deviation RMSE values greater than 0.40 m, whereas the EKF-based framework achieved lateral-deviation RMSE values of 0.031 m for straight-row trials and 0.034 m for headland U-turn trials relative to an independent taut-line reference under the tested GPS-denied maize-field conditions. These results demonstrate consistent row-following and navigation performance under the tested field conditions.