Traditional low-cost GPS installed on vehicles and other equipment has a tolerance of tens of meters. With the help of auxiliary devices and/or methodologies such as Differential GPSs (DGPSs), Assisted GPSs (AG-PSs), Real-Time Kinematic (RTK), computer vision and etc., the positioning accuracy of GPS receivers increased a lot, but a certain amount of cost increased at the same time. In this paper, we propose a new cooperative vehicular localization scheme for improving the accuracy of GPS fixes based on Vehicle-to-vehicle (V2V) communication. The proposed scheme first estimates the distances between neighboring vehicles using WLS-DD scheme [1] and uses a machine learning technique called Constrained Self-Organizing Map (C-SOM) with a set of GPS fixes collected over time to generate the final estimates of GPS locations with much lower errors. We present simulation results that demonstrate the superior performance of the proposed Constrained-SOM.