One of the most important functions for the safety of autonomous driving is the self-localisation of a moving platform. The GNSS (Global Navigation Satellite System) sensor is - compared to other sensors as, e.g., video, radar or camera - the only sensor that provides absolute positioning. Under clear sky conditions, GNSS positioning methods yield high level of accuracy and integrity. Poor signal propagation conditions make precise GNSS positioning challenging under urban conditions where multipath degrades the attainable accuracy and integrity. In the last decades, various correction or mitigation approaches for multipath errors have been developed and tested, but for carrier phase observations it is still the most critical error with magnitudes of up to a quarter of the wavelength. Since the carrier phase multipath error is location dependent, this property can be used to construct a correction map. In this paper, we report on the necessary steps to construct such a multipath correction map. The carrier phase error is a function of the multipath characteristics and the extra path delay. Hence, it is dependent on the distance to the reflection surface. Using a 3D building model, the geometry between the building and potentially multipath affected satellite signals is established to obtain one periodic cycle of the carrier phase error. A pseudo-kinematic experiment in an urban environment is planned and designed based on simulation studies of a ray-tracing algorithm to ensure ideal receiving properties for multipath signals in a realistic scenario. Two different approaches of carrier phase residuals will be compared to derive the multipath error from the observations. Firstly, double difference (DD) residuals are computed. Secondly, carrier phase residuals extracted from a positioning engine will be analysed. Finally, the developed functions are evaluated in the context of generating GNSS multipath correction maps for moving platforms