Abstract: In multi-robot simultaneous localization and mapping (MRSLAM), existing distributed pose optimization methods are limited in localization accuracy due to the lack of in-depth utilization of map point observation information. To address this issue, we propose a globally consistent adaptive distributed SLAM optimization algorithm under multi-view observations. The core innovation of the algorithm lies in the design of a tightly coupled optimization framework, which deeply integrates multi-view observation information on the basis of global pose consistency optimization. By minimizing the reprojection error of map points generated by distributed sparse bundle adjustment, the pose estimation between robots is further constrained to converge to the optimal solution, forming a virtuous closed-loop optimization. Meanwhile, the algorithm introduces adaptive weights to enhance robustness in noisy environments. Experiments on both simulated and real datasets show that, compared with algorithms such as SE-Sync, DGS, and DDF-SAM, the proposed algorithm can significantly reduce the average pose estimation error by 22% to 36% in various network scenarios, while maintaining a low iteration time per cycle, achieving an effective balance between accuracy and efficiency. This research provides a high-precision and highly robust solution for multi-robot cooperative SLAM optimization algorithms.