This paper systematically reviews the evolution and current state of artificial intelligence (AI) agents. It links the evolution of AI agents, from rule-based systems to learning-based autonomous systems, to technological advancements such as deep learning, reinforcement learning, and multi-agent coordination. Specifically, it highlights the challenges of designing and deploying integrated AI agents that seamlessly integrate cognition, planning, and interaction. It comprehensively analyzes various approaches, including cognitive science-inspired models, hierarchical reinforcement learning frameworks, and large-scale language model-based inference. Furthermore, it discusses ethical, safety, and interpretability issues associated with deploying AI agents in real-world environments, suggesting future directions for AI agent systems.
