The knee is a key weight-bearing and stabilizing joint in daily locomotion, supporting body weight, maintaining balance, and enabling flexible movement during level walking, ramp ascent and descent, and stair ascent and descent. With aging, older adults often experience declines in muscle strength and endurance, leading to weakened knee strength and insufficient body support, which can reduce daily mobility and quality of life while increasing fall risk. Against the backdrop of global population aging, developing knee assistive exoskeletons that can support older adults in real daily walking scenarios is highly important. Although various knee assistive exoskeletons have been developed, several practical challenges remain. Conventional motor-driven systems often add weight and inertia at the knee joint, increasing the burden on the wearer; pneumatic systems are compliant but limited by power supply, stability, and portability; and existing devices still struggle to recognize cyclic locomotion patterns such as level walking, ramps, and stairs in real outdoor environments. "Moreover, many assistance strategies are designed for specific locomotion modes and lack the flexibility to adapt to changing daily walking conditions." said the author Shisheng Zhang, a researcher at Shenzhen Institutes of Advanced Technology, "Therefore, there is a clear need for a lightweight, comfortable, safe, and flexible knee assistive exoskeleton that can accurately recognize multiple daily locomotion patterns and provide appropriate assistive torque in real time."
This study designed a single-source dual-drive flexible knee assistive exoskeleton (FKAE), using a single motor, clutch mechanism, and Bowden cables to provide time-shared assistance to the left and right knees. This design reduces the number of high-power motors and distal joint load, while a series elastic actuator improves interaction compliance and safety. The system consists mainly of a back-mounted drive control module, bilateral lower-limb exoskeletons, and a locomotion perception system. Four IMUs are mounted on the thigh and calf components to calculate knee angles in real time and extract gait-related features. For daily locomotion scenarios, including level walking, ramp ascent, ramp descent, stair ascent, and stair descent, the researchers proposed a dual-detection locomotion pattern recognition method combining fuzzy control and a finite state machine. Fuzzy control was used to recognize steady-state gait patterns, while the finite state machine detected transitions between different locomotion modes to reduce switching delay. Based on the recognized locomotion pattern, a finite-state time-shared assistive torque control strategy was further designed to generate assistive torque curves according to the real-time knee angle and provide knee-flexion assistance during level walking, ramp ascent, and stair ascent. Finally, 3 young participants and 3 older participants were recruited for outdoor locomotion recognition tests on level ground, ramps, and stairs, as well as indoor experiments using metabolic measurements and surface EMG signals to evaluate the effects of exoskeleton assistance on energy expenditure and muscle workload.
The experimental results showed that the single-source dual-drive flexible knee assistive exoskeleton could achieve high-precision locomotion pattern recognition in real walking scenarios while effectively reducing users' movement burden. First, the switching time between left- and right-knee assistance was 0.10 to 0.12 s, satisfying the gait-cycle requirements under different locomotion modes and supporting the feasibility of the single-motor time-shared driving strategy. In outdoor locomotion recognition experiments, the dual-detection strategy combining fuzzy control and a finite state machine accurately recognized daily locomotion modes, including level walking, stair ascent, stair descent, ramp ascent, and ramp descent, with an overall average recognition accuracy of 98.89%; stair ascent recognition reached 100%. During mode transitions, the finite state machine substantially reduced detection delay, and in some transitions from level walking to stairs or ramps, the system could even anticipate the change by about half a gait cycle. Energy-consumption experiments further showed that, compared with the zero-torque condition, exoskeleton assistance reduced metabolic cost by approximately 5.4% to 12.8% during level walking, 11.9% to 28.2% during ramp ascent, and 10.9% to 18.8% during stair ascent. It also reduced surface EMG signals in multiple lower-limb muscles, with some muscle activation reductions exceeding 60%. These preliminary results indicate that the system can provide effective knee assistance across multiple daily locomotion tasks, with good recognition accuracy, interaction safety, and load-reduction potential.
The significance of this work lies in proposing a lightweight and flexible knee assistive exoskeleton for elderly daily locomotion, integrating structural design, locomotion pattern recognition, and assistive torque control to adapt to changing real-life scenarios such as level walking, ramps, and stairs. The single-source dual-drive architecture uses one motor to provide time-shared assistance to both knees, reducing the number of high-power motors and overall system weight. The Bowden cable and series elastic structure reduce distal joint load while improving human–robot compliance and safety. At the same time, the locomotion recognition method combining fuzzy control and a finite state machine achieved high accuracy and low transition delay in both steady walking and mode-switching situations, providing a basis for timely and appropriate exoskeleton assistance. The experimental results also preliminarily showed that the finite-state time-shared assistance strategy could reduce metabolic cost and lower-limb muscle activation, suggesting the system's potential for supporting elderly daily mobility. However, this study remains an initial validation with a small number of participants. The single-source dual-drive structure cannot actuate both knees simultaneously and cannot cover all assistance demands across the full gait cycle. "In addition, the current recognition method mainly relies on limited kinematic features, and its robustness under complex terrain, turning, nonperiodic gait, and individual variability still needs further validation. Future work could expand testing in older populations, integrate multisource sensing such as foot pressure or EMG, and introduce adaptive control and human-in-the-loop optimization to improve personalized assistance and real-world applicability." said Shisheng Zhang.
Authors of the paper include Shisheng Zhang, Yang Zhang, Yanzong Xu, Jinke Li, Yuquan Leng, and Xinyu Wu.
This work was supported in part by the National Natural Science Foundation of China (grants 62125307, 52175272, and 62403452), the Guangdong Basic and Applied Basic Research Foundation (grants 2024B1515020008 and 2023B1515130007), the Shenzhen Science and Technology Program (grants RCYX20231211090345058, JCYJ20220530114809021, and KCXFZ20230731093059012), and the Natural Science Foundation of Top Talent of SZTU (grant GDRC202328).
The paper, "A Single-Source Dual-Drive Flexible Knee Assistive Exoskeleton for Elderly Daily Locomotion" was published in the journal Cyborg and Bionic Systems on Aug 11, 2026, at https://doi.org/10.34133/cbsystems.0576.
