Vision-and-Language Navigation has emerged as a key capability for next-generation unmanned systems, enabling embodied agents to interpret natural language instructions and autonomously operate in complex, real-world environments. This paper systematically reviews the evolution of VLN, from early indoor, graph-based settings to recent advances in outdoor and aerial scenarios. We highlight the pivotal role of foundation models, including large-scale vision–language models (VLMs) and large language models (LLMs), in enhancing open-vocabulary perception, compositional reasoning, and generalization, thereby enabling robust zero-shot navigation and collaborative multi-agent systems. The review covers major benchmarks, representative methodologies, and state-of-the-art systems, with a particular focus on the challenges and opportunities arising from dynamic urban and aerial environments. Finally, we discuss recent key innovations and persistent challenges in the field, and outline future directions for the development of intelligent, language-guided autonomous agents.