Abstract: Objectives To reduce the accumulated positioning error of autonomous underwater vehicles in underwater heterogeneous formations, a USV-AUV cooperative bathymetric simultaneous localization and mapping method based on a distributed factor graph is proposed. Methods Real-time pose constraints between the USV and multiple AUVs are established using ultra-short baseline acoustic positioning. The SINS/DVL odometry constraints, USBL positioning constraints and terrain-matching loop constraints are represented in a factor graph to achieve global optimization of AUV poses and bathymetric submaps. To reduce the data transmission load over low-bandwidth underwater acoustic channels, a Morton-code optimized octree and arithmetic coding method is adopted for lossless compression of bathymetric submaps. A three-stage terrain matching strategy combining globally optimal coarse registration, generalized iterative closest point fine registration and pairwise consistency maximization is further designed to suppress false loop closures. Results Simulation results with one USV and three AUVs show that, compared with the EKF-based cooperative SLAM method, the proposed method improves AUV positioning accuracy by 22%, increases global mapping consistency by 16.05%, and reduces the single-frame bathymetric submap data size from the MB level to the KB level.Conclusions The proposed method can improve the positioning and mapping accuracy of USV-AUV heterogeneous formations while reducing underwater acoustic communication load, providing a feasible technical route for cooperative bathymetric mapping.