Situated within human society, embodied agents are continuously exposed to diverse streams of information, ranging from visual observations to natural language interactions. A central challenge is enabling them to learn from and effectively leverage this information over extended periods. To address this, we introduce Ella, an embodied lifelong learning agent designed to accumulate experiences and acquire knowledge across hours of social interaction in a 3D open world. At the core of Ella’s capabilities is a structured, non-parametric, long-term multi-modal memory system that stores, updates, and retrieves information effectively. It consists of a name-centric semantic memory for organizing acquired knowledge and a spatiotemporal episodic memory for capturing multimodal experiences. By integrating foundation models with this non-parametric memory system, Ella retrieves relevant information for decision-making, plans daily activities, builds social relationships, and evolves autonomously while coexisting with other intelligent beings in the open world. We conduct capability-oriented evaluations in a dynamic 3D open world where 15 agents engage in social activities for days and are assessed with a suite of unseen controlled evaluations. Experimental results show that Ella can influence, lead, and cooperate with other agents well to achieve goals, showcasing its ability to learn effectively through observation and social interaction. Our findings highlight the transformative potential of combining non-parametric memory systems with foundation models for advancing embodied intelligence.
Ella: Embodied Lifelong Learning Agent
We build long-term memory in two forms:
(a) name-centric semantic memory organizes the knowledge in a name-centric graph including a hierarchical scene graph serving as the spatial memory;
(b) spatiotemporal episodic memory stores the experience as a series of events consisting of
time,
location, and multimodal contents.
(c) Ella first generates a daily schedule according to the knowledge and experiences retrieved from the long-term memory,
(d) then updates the memory based on visual observations of the environment, and
(e) social interactions with other agents and
(f) makes reactions accordingly including
(f1) revising the schedule,
(f2) interacting with the environment,
(f3) and engaging in a conversation.