RoDyn-SLAM is a novel dense RGB-D SLAM framework that leverages neural radiance fields (NeRF) to robustly track camera motion in dynamic environments. The key contributions of RoDyn-SLAM include: 1. **Motion Mask Generation**: A method to filter out invalid sampled rays by fusing optical flow masks and semantic masks, enhancing the precision of motion masks. 2. **Pose Optimization**: A divide-and-conquer algorithm that distinguishes between keyframes and non-keyframes, using an edge warp loss to improve geometry constraints between adjacent frames. 3. **Performance**: Extensive experiments on challenging datasets show that RoDyn-SLAM achieves state-of-the-art performance in both accuracy and robustness compared to recent neural RGB-D methods. The method addresses the limitations of traditional visual SLAM methods in dynamic environments by improving the accuracy and robustness of pose estimation. The implementation of RoDyn-SLAM will be open-sourced to benefit the community.RoDyn-SLAM is a novel dense RGB-D SLAM framework that leverages neural radiance fields (NeRF) to robustly track camera motion in dynamic environments. The key contributions of RoDyn-SLAM include: 1. **Motion Mask Generation**: A method to filter out invalid sampled rays by fusing optical flow masks and semantic masks, enhancing the precision of motion masks. 2. **Pose Optimization**: A divide-and-conquer algorithm that distinguishes between keyframes and non-keyframes, using an edge warp loss to improve geometry constraints between adjacent frames. 3. **Performance**: Extensive experiments on challenging datasets show that RoDyn-SLAM achieves state-of-the-art performance in both accuracy and robustness compared to recent neural RGB-D methods. The method addresses the limitations of traditional visual SLAM methods in dynamic environments by improving the accuracy and robustness of pose estimation. The implementation of RoDyn-SLAM will be open-sourced to benefit the community.