Multi-robot SLAM is a technology of multi-robot simultaneous localization and mapping cooperatively in unknown environment. We use information filters algorithm to solve multi-robot SLAM problem. Information filers use information matrix to describe the SLAM uncertain information, and multi-robot SLAM information matrix contains all robots poses and all the features observations which observed by all robots. We set up the multi-robot models, motion models and observe models, and set up feature-environment to simulation in computer. We set threshold to divide the features which are observed by robot k on current time, one part of features are used to relocalization as active maps, and the other features are used to update information matrix and information state. The simulation data shows that information filters can keep the accuracy of the robots poses, and keep the motion update and the observation update in the constant range, the algorithm improve the effective of multi-robot SLAM.