Abstract：Large model-driven embodied intelligent robot navigation technology is a frontier field at the intersection of artificial intelligence and robotics. It focuses on leveraging the natural language processing， multimodal information fusion， and logical reasoning capabilities of large models to endow robots with the ability to achieve autonomous environmental perception， semantic instruction understanding， and precise navigation decision-making in dynamic and complex environments. This approach breaks through the bottlenecks of traditional navigation methods in terms of adaptability， generalization， and human-robot interaction in unstructured scenarios. In recent years， with the technological breakthroughs in large language models and multimodal models， the navigation methods for embodied intelligent mobile robots have been evolving from the reinforcement learning paradigm to the general intelligence paradigm， demonstrating significant application potential in scenarios such as industrial automation， intelligent services， and disaster relief. This paper systematically reviews the technological evolution and current research status in this field. Firstly， it reviews the theoretical origins of embodied intelligence and the development history of navigation technology， and analyses the integration paths between artificial intelligence technology and embodied intelligent navigation. Secondly， from an architectural perspective， the current mainstream embodied intelligent navigation methods are categorized into two core paradigms： end-to-end models and hierarchical models， with their technical principles， representative models， and application scenarios expounded respectively. Subsequently， the common datasets and mainstream evaluation metrics for embodied navigation tasks are introduced. Finally， combined with the current technical bottlenecks， future research directions are pointed out. This paper aims to provide systematic technical references for researchers in this field and promote the transition of large model-driven embodied intelligent robot navigation from theoretical research to practical applications.
2025, Vol. 59 
