Discovering approximate expressions of GPS geometric dilution of precision using genetic programming

Global Positioning System (GPS) has been used extensively in various fields. Geometric Dilution of Precision (GDOP) is an indicator showing how well the constellation of GPS satellites is organized geometrically, so as a reliability indicator presenting the GPS positioning accuracy. Traditional methods for calculating GPS GDOP need to solve the measurement equations where involve complicated matrix transformation and inversion. Some studies rephrase the calculation of GPS GDOP a regression problem and employ “black-boxed” machine learning methods for problem solving. However, the regression models obtained from such methods lack of expressivity for describing the relationships among variables. Making the structures of GDOP expressions visible is valuable because they can be further studied or tailored for specific GPS applications. This study employs the technique of genetic programming (GP) for the regression of GPS GDOP. The performance of GP working with various operators and parameter settings is studied and discussed. The experimental results show that GP generates precise models with better expressivity for GPS GDOP than other methods.