Abstract: Both uncertainty about the pose and the environment maps,simultaneous localization and map building（SLAM） becomes so complex in unknown environments.So this paper introduces a local optimal parameter approach：first to build local optimal environment maps by local optimal robot pose,then to find out local optimal robot pose by local optimal environment maps.Localization and map building are performed recursively until the task is completely finished.The experiment results indicate that the SLAM finished by the modified SLAM based on Particle Filtering（PF-SLAM） is more accurate than the normal PF-SLAM.It gives a new method for robot SLAM in unknown outside environments.