An improved multi-robots cooperative online FastSLAM algorithm based on auxiliary measurements is proposed for the key problems of FastSLAM 1.0 and 2.0. In FastSLAM 1.0, the cumulated errors are presented for the uncertain localization of the robot itself without measurements. Although FastSLAM 2.0 takes the measurements into account, additional computational complexity is introduced. In the proposed algorithm, two robots are required. The leader robot is responsible for accomplishing the task of SLAM, and the auxiliary robot is responsible for providing the leader robot with static relative measurements. The method of cooperative auxiliary measurement provides the SLAM executor with more accurate localization information in FastSLAM 1.0 and avoids the additional computational complexity in FastSLAM 2.0. Experimental results show that the algorithm can achieve higher accuracy and feasibility. Moreover, the method is of high practical value.