Abstract: pathway to overcome the adaptability bottlenecks of industrial robots in challenging environments. This study provides a systematic review of multi-modal perception technology and its typical applications in complex scenes, and focuses on the assembly tasks of unstructured and irregular parts. An experimental platform was constructed to conduct comparative tests. The results indicate that the systems incorporating multi-modal perception technology significantly outperform the traditional solutions in terms of recognition accuracy，assembly success rate，and operational stability，thereby validating their engineering practicality and promotion value in complex industrial settings.