Multisensory Continual Learning: Adapting Pretrained Visuomotor Policies to Force

This research addresses the challenge of adapting pretrained robot policies to new sensory modalities like force and tactile sensing. The proposed MuSe method integrates limited multisensory data into vision-only policies through multi-stage fusion, future prediction, and experience replay. Evaluated on real-world tasks, MuSe shows strong performance in contact-rich scenarios while preserving or improving original task performance.