CloudEdgeSLAM (IEEE TVT 2023)

Learning-based SLAM methods (e.g., DROID-SLAM) are often too heavy for edge robots, while model-based SLAM (e.g., ORB-SLAM) can fail in challenging conditions.

We propose a cloud-edge collaborative system: a cloud-side learning-based SLAM cooperates with an edge-side model-based SLAM, using downsampling and submap merging strategies to improve robustness and success rate.

• Paper: https://ieeexplore.ieee.org/abstract/document/10264105