This paper introduces a laser-scanner measurement model using the statistic of laser-scanner data collected in advance for a mobile robot localization. In autonomous navigation, robots usually run based on self position on a map, and laser-scanners are useful sensors for localization. However, in human living environments like urban areas and parks, laser-scanner data is unstable due to moving objects and natural objects, and it is difficult to obtain landmarks like fixed objects. Therefore, our method make a map using statistics of laser-scanner data and calculates the laser-scanner measurement model based on statistics. Our method is applied to Monte Carlo localization/particle filter. Because the map makes possible to use the frequency and distribution of laser-scanner data for localization, our method allows a robust localization for unstable laser scanner data. In extensive experiments, our method presented an accurate localization and a robot using our method ran stably in actual sidewalks.