Intelligent vehicles have gained increasing popularity in recent years as traffic safety and efficiency have become the major challenges faced by automotive industry. Vehicle positioning system, such as GPS, plays more and more important role on intelligent or autonomous driving. Intelligent vehicle technologies have been developed and tested mainly based on intensive field experiment under various driving scenarios. However, the large variation, uncertainty and complexity of the driving environment, including buildings, traffic and weather conditions have posed great challenges on test repeatability and system robustness. This paper proposes a GPS model considering software-centered observation errors. The focus of the research is on its error to reflect the real signals from GPS measurement. A three-dimensional position of the twenty-four GPS satellites is first calculated according to the ephemerides, and the visible satellites are selected based on the position with respect to the satellites. Then the distance between them is calculated as the ideal signal, which is further contaminated with Gaussian white noise to simulate the pseudo range that stems from various noise sources from hardware and environmental factors. Simulation has been conducted and the results are correlated well with real signals measured by an on-board GPS receiver.