Design of a Low-cost GPS/magnetometer System for Land-based Navigation: Integration and Autocalibration Algorithms

The land-based navigation, paying attention to precision farming, is the research topic: the final purpose is the design and development of a guidance-aided system focusing on a low-cost GPS receiver able to provide a pseudorange-based solution only. Specific tests have been carried out to reproduce the trajectories followed by the vehicle in agricultural applications, whose accuracy target is typically 1 m. Results show that the investigated low-cost receiver is affected by a drift in time which is mainly detected while turning and causing a deviation from the optimal reference solution. Thus, the goal is to correct this behavior because the deviation accumulates during time and causes a not optimal treatment of the field (waste of material and money). Paying attention to the cost of the system, a new idea is proposed: the integration between the low-cost GPS with a magnetometer/digital compass. A dedicated algorithm has been also implemented, taking the heading provided by th e magnetometer and using it to correct the deviation in turns. Unluckily a magnetometer is deeply influenced by ferrous materials and the sensor is supposed to be installed on the vehicle, which is mainly made by metal. As a consequence, the sensed measurements are affected by a deviation from the actual magnetic field. Those disturbances need to be properly reduced by an autocalibration procedure. A new approach for the autocalibration problem has been developed and implemented; then the comparison with respect to the traditional method has been also performed in order to test and validate the new idea. A comprehensive and detailed description of all the algorithms will be produced concerning both the sensors integration (GPS and magnetometer) along with the magnetometer autocalibration. Particular attention will be focused on results and performances of the autocalibration procedure, which appears to provide very interesting results. The new approach, which is simply based on the covariance matrix, appears to be more successful than the ...