Integrating LLMs, Knowledge, and Reasoning for General Embodied AI.

The pursuit of general embodied intelligence (GEI) is gaining momentum, driven by the convergence of large language models (LLMs), structured knowledge bases, and sophisticated reasoning abilities. This research provides a comprehensive overview of how these components can be integrated to create the next generation of AI agents capable of interacting with and understanding physical environments. The paper analyzes current LLM architectures, training methods, and inference mechanisms, highlighting their interaction with external knowledge sources and logical reasoning frameworks. It also explores existing embodied intelligence paradigms where agents learn and act within physical settings. A conceptual framework is introduced to illustrate the synergistic relationship between LLMs, knowledge bases, reasoning, and embodiment, serving as a guiding model for perception, reasoning, and action. To advance towards GEI, the authors pinpoint five critical challenges: efficient LLM deployment, seamless closed-loop knowledge integration, hybrid symbolic-neural reasoning, robust perception-action grounding, and continuous learning. This work offers a strategic roadmap for developing highly adaptive, multimodal AI agents that can operate effectively in complex and dynamic real-world scenarios.