Based on the essential characteristics of driving behavior, this paper constructs a behavior state modeling and semantic annotation system, proposes a behavior-task mapping rule generation model, designs a rhythm regulation and channel management mechanism, and finally forms a safety policy execution architecture with hierarchical intervention capability. Under this framework, the interactive system can dynamically adapt to driving operation differences, adjust the rhythm of information output, manage multi-modal channels, and realize pre-intervention control for high-risk behaviors.