**5. Conclusions**

This manuscript uses ORB-based feature point re-identification to improve the positioning accuracy of all unmanned platforms in a cluster by fusing the local maps of each unmanned platform with new position constraints and global GNSS positioning information. Through a centralized collaborative positioning service, this can provide low latency positioning information for subsequent collaborative path planning and task allocation algorithms. The co-location algorithm proposed in this manuscript has better accuracy in both GNSS-challenged and GNSS-denied modes than the ORB-SLAM3 algorithm running on a single platform. The two-stage position estimation method used can also be combined with other positioning sensors such as UWBs and barometers in addition to the GNSS global positioning information applied in the manuscript. Current techniques for the localization of air-ground unmanned clusters in complex environments present new demands in the direction of visual front-ends, optimization methods and multi-sensor fusion. The approach proposed in this manuscript can be extended to other unmanned clusters in areas such as UAVs, logistics, agriculture and military. In the future, based on the current results, our subsequent work will further investigate the impact of different vision front-end techniques on location re-identification, a key aspect that critically affects map fusion and closed-loop detection. We will consider the use of feature detection and matching techniques based on deep learning or point and line features to improve the robustness of the algorithmic framework and to improve the localization accuracy based on this. Finally, our proposed approach in this manuscript has high computational resource requirements for map fusion and global optimization, and we will aim to mitigate the computational power and communication bandwidth required for global map maintenance on a server.

**Author Contributions:** Conceptualization, H.X.; methodology, H.X.; software, H.X.; validation, Y.L. and S.Y.; formal analysis, C.W.; resources, Y.B.; data curation, C.W.; writing—original draft preparation, H.X.; writing—review and editing, C.J.; visualization, W.L.; supervision, W.L.; project administration, Y.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

