The interaction force between the patient and lower limb exoskeleton determines the safety and comfort of patients in the process of rehabilitation. In order to provide comfortable and safe rehabilitation training, in this paper, the coupling relationship between patients and lower limb exoskeleton was considered, the man-machine coupling dynamic equation was established based on Lagrange method, and the man-machine coupling model was built on ADAMS platform. Traditional impedance control can adjust the relationship between force and position in real time, which is very suitable for man-machine coupling control system. However, it is unable to adjust the parameters of the controller in real time. On this basis, a torque feedback control method based on RBF fuzzy adaptive variable impedance was proposed, which could adjust the parameters of impedance controller in real time. Finally, the united simulation was carried out on the MATLAB-ADAMS platform. The simulation results show that compared with the traditional impedance control, the control method proposed in this paper reduces the joint angle tracking error between the lower limb exoskeleton and the human body by 44.2%, and the consistency between the change trend of the joint torque of the lower limb exoskeleton and the change trend of the joint torque of the human body is greatly improved, The control method proposed in this paper can improve the compliance and coordination of man-machine coupling system.