In order to improve the current problems of lower limb rehabilitation robots for patients after orthopedic surgery, such as limited types and weak interaction, a new type of sitting and lying lower limb rehabilitation robot based on human dynamics modeling and tension and pressure sensors to achieve human-computer interaction is designed. The system adjusts the training angle of the hip, knee and ankle joints based on real-time monitoring of human-computer interaction, verifies the established mathematical model and compares the actual collected joint angles, and concludes that the mathematical model has the same trend of torque changes in Matlab fitting and Adams simulation, and the maximum average variance does not exceed 0.81N/m. The experimental comparison of the joint angles of the fully passive and intelligent passive states has been significantly adjusted, which proves that the system increases the patient’s active participation.