Distributed cooperative localization can improve the localization accuracy for robots in absolute measurements intermittent or denied scenarios. Existing distributed cooperative localization methods often use fixed process noise covariance matrices(PNCM) for extended Kalman filter execution, whose localization accuracy deteriorates when the PNCM is unknown or time-varying. This paper proposes an adaptive cooperative localization algorithm based on the Sage-Husa adaptive filter to address the unknown process noise statistics problem for the 2-D multi-robot system. The effectiveness and superiority of the proposed algorithm is proved in simulations.