In the relative technical researches of the mobile robot, the ability to achieve autonomous navigation represents the intelligent level of the robots, and the localization is the key to achieve autonomous navigation for the Monte Carlo localization algorithm is one of the common methods used in the researches of the mobile robot’s self-positioning technology, it has been widely used because of its parallel, easy to implement and can effectively deal with the characteristics of nonlinear problems. However, the traditional Monte Carlo localization algorithm has some problems, such as particles running out and particle degrading, which can not guarantee the accuracy of the robot localization. Some improved algorithms can ensure the accuracy, but they increase the complexity of the algorithms and are not easy to achieve. The purpose of this paper is to improve the traditional Monte Carlo localization algorithm to propose a location algorithm which is easy to implement and can effectively deal with non-linear problem. Firstly, this paper analyzed and established the mobile robot localization related models such as coordinate system model, motion model, the sensor observation model, noise model. Secondl