Study On Navigation System For Off-Road Vehicle Guidance Based On GPS And Sensor Technology

The objective of this dissertation is to integrate inertial sensors, global positioning system and dynamic off-road vehicle model for supporting accurate tractor automated guidance. The main research has two parts. One is to create a dynamic model for estimating trajectories of a tractor-implement system traveling in the field to provide necessary information, the other is to develop a positioning system using low-cost GPS and solid-state inertial sensors to provide improved positioning accuracy in sub-meter range for agricultural vehicles use in the field.The main challenge of developing the dynamic steering model is to describe the tractor-implement steering dynamics system without the loss of generality. Another challenge is the accuracy of the trajectory estimation which is strongly affected by the determination of the cornering stiffness of the tractor. Because the value of cornering stiffness is greatly influenced by the characteristics of soil and wheel interactions, in this model analysis it was determined using the specific procedures.To model a tractor-implement system with sufficient details but without loss of generality, this tractor system model was developed based on a John Deere 7700 tractor. This model, consisting of a bicycle model of the tractor sub-system, has been implemented in the MATLAB environment. The simulation results from this MATLAB model was validated through field navigation tests. The accuracy of the trajectory estimation is strongly affected by the determination of the cornering stiffness of the tractor. Because its value is greatly affected by the characteristics of soil and wheel interactions, the tractor cornering stiffness in this simulation analysis was determined using the following procedures: (1) a series of circling tests of the investigated tractor system was performed at different speeds on the testing fields; (2) the tractor center-of-gravity trajectory was recorded using a RTK-DPGS; (3) an appropriate cornering stiffness was identified during simulation analysis using the MATLAB model based on the recorded trajectory data; and (4) the obtained cornering stiffness was used in simulation analyses for various navigation operations in the field of interest. The analysis on field validation test results indicated that the developed tractor system could accurately estimate wheel trajectories of a tractor system while operating in agricultural fields at various speeds. The results also indicated that the developed system could accurately determine tractor velocity and steering angle whilethe tractor operates in curve fields.The challenge of developing the low-cost navigation system is discovering how to efficiently integrate the available data from the GPS and IIS into an accurate positioning solution. To meet the requirements of most tasks for off-road vehicle and precision farming applications, this navigation system should reach positioning accuracy in the sub-meter range. In addition, this navigation system should be able to provide reliable positioning signals at a sufficiently high update rate even during GPS signal outage.To effectively integrate the data from the GPS and IIS, a Kalman filter, based on a position-velocity-attitude (PVA) model of the vehicle, is developed for the GPS/IIS fusion to reduce the positioning error, increase the update rate of the fusion system, and bridge the GPS signal outage. With this fusion algorithm, the navigation system can accurately estimate vehicle position information based on the redundant data from various sensors. In addition, in order to reduce the computational intensity of the Kalman filter, this dissertation also proposes a different GPS/IIS fusion algorithm, which is based on an IIS error model and a complementary Kalman filter.To evaluate this low-cost navigation system and associated fusion algorithms, validation tests are performed using an experimental system that is designed to acquire real-time data and to perform data fusion. These tests take place at three different test sites that include flat and uneven terrain, with...