The localization of vehicles in complex and dynamic urban environments is a crucial area of research, especially given that autonomous navigation technology plays a key role in addressing localization challenges. When Kalman-based filters are used for localization, sensor observation outliers caused by urban environmental factors introduce errors that are difficult to correct. These errors significantly reduce localization accuracy and can even lead to divergence in estimation. Moreover, modern vehicle localization algorithms often encounter computational bottlenecks, limiting their overall efficiency. Building on existing research, we propose an optimized and robust localization algorithm based on the maximum generalized Versoria criterion (MVC). Numerical simulations in nonlinear vehicle state estimation systems confirm the proposed method’s effectiveness across various types of nonlinear measurements.