Yong Wang, Boyu Shen, Xinwei Zhang, Yizheng Li, Keyuan Luo
This survey presents a taxonomy of memory mechanisms for Vision-Language-Action models, categorizing existing approaches into five paradigms. It compares representative systems and discusses key challenges in building full-lifecycle memory systems. The survey provides insights into developing lifelong embodied intelligence.
The development of effective memory mechanisms is crucial for enabling robots to perform long-horizon tasks and make adaptive decisions. This research has practical significance in areas like robotics and artificial intelligence. It can lead to the creation of more intelligent and autonomous systems.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.