Context-aware embodied intelligence: from environmental understanding to autonomous decision-making

In recent years, embodied intelligence has emerged as a significant research direction in artificial intelligence, aiming to enable an agent to autonomously achieve intelligent behaviors in complex environments through the integration of perception, interaction, and decision-making capabilities. This research area not only focuses on the computational processes of internal algorithmic models within the agent but also emphasizes the physical structure of the agent and its real-time interaction with the external environment as core elements to the emergence of intelligence. Context awareness introduces a novel research perspective to embodied intelligence, serving as a key driving force that empowers the agent to understand environmental semantics, adapt to dynamic changes, and autonomously execute efficient decision-making. On this basis, this paper provides a systematic review of recent advancements in embodied intelligence, focusing on aspects such as environmental understanding, cognitive reasoning, and autonomous decision-making. In addressing the challenges of real-time semantic understanding of multi-source sensory data and continuous decision-making in long-term tasks, this paper further analyzes how the emerging paradigm of context-aware embodied intelligence, centered on perception-action closed-loop frameworks, relies on effective context-aware modeling and reasoning to adapt autonomously to dynamic environments and task constraints. This framework enables the agent to achieve more efficient and contextually aligned task execution in real-world scenarios. Additionally, this paper explores remaining challenges and outlines potential future research directions in embodied intelligence.