This paper presents a new method of calibrating the alignment error between GPS (Global Positioning System) and LiDAR (Light Detection And Ranging) for a low cost MMS (Mobile Mapping System) during initial installation. Through a detailed mathematical analysis, the calibration problem is transformed into searching the minimum value of a nonlinear multivariable objective function. Thus, a new method is proposed as a solution to this problem. That is the Nelder-Mead simplex direct search algorithm. After calibration, the average reduction of point cloud data relative error (The distance between two points witch obtain from scanning one point in the space twice by LiDAR) is 3.85cm(within 40m range from LiDAR). The maximum relative error reduces from 8.55cm to 4.76cm. It is less than 5cm and meets the accuracy requirement in three-dimensional urban mapping.