Autonomous agricultural robots play an essential role in agricultural mechanization. Self-reliable navigation of these robots is a challenging task. Generally, Global Navigation Satellite Systems (GNSS) based approach, combined with other sensors, is mainly utilized for robot navigation. However, these navigation methods encounter trouble during GNSS signal outages and require a stable navigation solution. This paper proposes a stable navigational algorithm for an orchard robot based on the model-based design that works in a GNSS denied environment. The designed model takes the input driving coordinates and vehicle state from wheel and steering encoders and translates them into vehicle navigational command. The navigation model is built in MATLAB/Simulink environment. The dynamic behavior of the model is neglected, considering that the robot is a slow-moving vehicle due to its application requirement. For robot navigation, the control command given by the controller is divided into velocity and steering control that follows a defined path at a given velocity and steering rate. The developed system is verified both in real and simulation environments. The control algorithm is verified using a mobile robot (non-holonomic) in an outdoor environment. It is confirmed that the robot is able to navigate satisfactorily in the given environment with normalized root mean square difference error within the range of 0.2–0.4 for steering offset at turning and 0.05 degrees during straight travel. Similarly, the model correlation lies close to 0.99 compared to the steering input angle. The proposed navigational model gives a realistic approach for vehicle navigation with the least sensor interaction.