When humans interact with robots at “zero distance”, human-robot interaction faces challenging security issues such as the possibility of bodily harm from the robots - a problem that can no longer be solved by traditional safety measures based on human-robot distance or physical parameter space constraints. Therefore, this paper introduces a new security approach, human-robot cooperative safety model, whose main innovations are: designing a human-robot cooperative safety model based on the understanding of the user’s intention, and realizing a self-enclosed human-robot cooperative assembly system based on multimodal perception. The experimental results show that the method can effectively avoid the danger of personal safety caused by misrecognition and misinterpretation of the robot in the interaction process, improve the intention understanding ability in the interaction system, reduce the user’s fear of the robot, and enable the user to complete the human-robot interaction session more realistically and smoothly.