Abstract: This paper proposes a logical architecture for intelligent radar systems, establishing three core mechanisms: a hierarchical perception-cognition processing analysis mechanism, a decision-action management scheduling mechanism, and a human-machine interaction closed-loop feedback mechanism. The perception-cognition mechanism achieves precise environmental awareness through signal reception, real-time processing, multi-source data fusion, and battlefield situational analysis. The decision-action management scheduling mechanism significantly enhances autonomous decision-making efficiency in complex scenarios via task planning, dynamic waveform scheduling, and self-optimized resource allocation. The human-machine interaction closed-loop feedback mechanism constructs an  enhanced feedback loop of "perception-decision-correction" through multimodal interaction and environmental adaptive optimization, significantly improving system anti-jamming robustness. Deeply integrating reinforcement learning and deep learning technologies, this architecture demonstrates superior performance in dynamic parameter adjustment, intelligent anti-jamming capabilities, and multimodal collaborative efficiency. It forms a closed-loop system of "autonomous perception-learning optimization-autonomous decision-making" in complex electromagnetic environments, combining adaptive optimization capabilities with proactive decision-making advantages. This research provides critical technical pathways for next-generation radar development.