Abstract: Faced with complex and dynamic environments as well as adversarial targets, existing radar detection methods generally fall into two categories: upgrading hardware to build a "perceptual information" hardware system by acquiring "more data", or improving software to transform "information data" into "knowledge representation" for building "thinking model" software systems. However, these levels of intelligence remain at the "sensing + AI" stage. This paper introduces the rapidly emerging concept of "embodied intelligence" from the field of robotics into the domain of radar detection and sensing. Through conceptual connotation analysis, system architecture design, key technology exploration, and application scenario prospects, this paper provides guidance for the development of "embodied intelligent radar" and promotes radar detection to enter the stage of "AI + sensing".