Abstract: To address the insufficient human–robot interaction adaptability in current lower-limb rehabilitation exoskeletons, this study proposes a lower-limb assistive exoskeleton featuring active actuation at the hip and knee joints and passive following at the ankle joint. A rigid self-adaptive human–robot interface is adopted to provide both effective assistive torque and structural compliance, aiming to deliver gait assistance for individuals with muscle weakness while improving wearing comfort.First, a bionic geometric structure of the exoskeleton was constructed based on human lower-limb biomechanics, and passive degrees of freedom in the attachment components were designed using a serial-chain topology. The forward kinematics model of the exoskeleton was then established and verified using the D–H method. Finally, a human–exoskeleton coupled model was built on the OpenSim platform, in which the human model was configured to two weakened muscle strength conditions (80% and 60% of the maximal isometric force). The assistive performance and interaction characteristics of the designed exoskeleton were evaluated through changes in two key indicators—overall metabolic consumption and human–robot interaction forces—as well as variations in hip and knee flexor–extensor muscle forces.The results show that the proposed exoskeleton effectively reduces human–robot interaction forces, with peak interaction forces at the thigh decreasing from 70 N to 20 N and those at the shank decreasing from 150 N to 30 N. The overall metabolic cost of the wearer was reduced by 13.8%–15.4%, and the muscle force outputs of major hip and knee muscle groups markedly decreased. These findings validate the rationality of the design and demonstrate its performance in adaptive human–robot interaction assistance, highlighting its application potential in rehabilitation training and gait assistance for individuals with muscle weakness.