GPS and odometer data fusion for outdoor robots continuous positioning

Present work describes an approximation to obtain the best estimation of the position of the outdoor robot ROJO, a low cost lawnmower to perform unmanned precision agriculture task such are the spraying of pesticides in horticulture. For continuous location of ROJO, two redundant sensors have been installed onboard: a DGPS submetric precision model and an odometric system. DGPS system will allow an absolute positioning of the vehicle in the field, but GPS failures in the reception of the signals due to obstacles and electrical and meteorological disturbance, lead us to the integration of the odometric system. Thus, a robust odometer based upon magnetic strip sensors has been designed and integrated in the vehicle. These sensors continuosly deliver the position of the vehicle relative to its initial position, complementing the DGPS blindness periods. They give an approximated location of the vehicle in the field that can be in turn conveniently updated and corrected by the DGPS. Thus, to provided the best estimation, a fusion algorithm has been proposed and proved, wherein the best estimation is calculated as the maximum value of the join probability function obtained from both position estimation of the onboard sensors. Some results are presented to show the performance of the proposed sensor fusion technique.