Abstract: This paper presents a novel variable virtual impedance control (VVIC) strategy which can adapt HEI to different pilots with a virtual impedance controller. The controller is model-based with a virtual impedance which models HEI between the pilot and the exoskeleton. To adapt different pilots with different HEI, a reinforcement learning method based on policy improvement and path integrals (PI²) is employed to adjust and optimize parameters of virtual impedance. We demonstrate the efficiency of the proposed VVIC strategy on a single degree-of-freedom (DOF) exoskeleton platform as well as a human-powered augmentation lower exoskeleton (HUALEX) system. Experimental results indicate that the proposed VVIC strategy is able to adapt HEI to different pilots and outperforms traditional model-based control strategies in terms of interaction forces.