A Continual Learning Survey for Robotics in Multimodal Scenarios

Continual learning enables robots to learn new things without forgetting what they have learned previously. However, it is still a great challenge for robots to process continuous multimodal data streams in a real-world scenario. In this paper, we investigate many multimodal continual learning methods to explore the future research direction in this field. In conclusion, although certain achievements have been made in current research, further exploration is still needed in the future in directions such as theoretical deepening, technological breakthroughs, and application expansion. The research achievements in this field are of great significance for promoting the wide application of robots in complex environments.