A review of artificial intelligence research in the International Journal of Hydromechatronics suggests that combining large language models (LLMs), structured knowledge systems, and physical agents could help create machines capable of understanding, reasoning, and acting in complex environments.
The study examines how LLMs, AI systems trained on extensive text data, might be integrated with knowledge bases, which organise information about the world, and reasoning mechanisms that can produce logical conclusions. The researchers propose a conceptual framework for general embodied intelligence (GEI) and show how these elements might interact.
Current LLMs, familiar to many people as ChatGPT, Gemini, Claude, and others, have strong language abilities but lack direct connection to the physical world. By contrast, embodied AI systems, such as robots, can perform tasks like navigation and object handling in the physical world but commonly do not have inbuilt complex reasoning and adaptation systems.
The researchers suggest that combining these two areas of technology could allow agents to use language for planning, memory, and decision-making while working with real-world conditions. However, challenges remain, including the symbol grounding problem, which boils down to the inherent difficulty in converting abstract language into physical actions.
The review identifies five priorities for progress: efficient AI deployment, integration of updated knowledge sources, improved neural and symbolic reasoning, stronger links between perception and action, and continual learning. If these are implemented successfully, then applications in healthcare support, adaptable manufacturing, and robotics for domestic and even hazardous environments might open up.
Yuan, F., Huang, X., Wang, L., Ding, J., Tian, Z., Wang, Y., Gu, S., Funabora, Y., Peng, Y. and Mao, Z. (2026) ‘Towards general embodied intelligence: integrating large language models, knowledge bases, and reasoning capabilities to build the next generation of AI agents’, Int. J. Hydromechatronics, Vol. 9, No. 2, pp.250–316.
News media may use this press release as source material, in whole or in part, provided the content is not materially misrepresented. A link back to the original article is appreciated.
