Data-driven controllers have enabled lower-limb robotic exoskeletons to adapt assistance to the user across many tasks. A leading example is biological torque control, which estimates the user's own joint moment from onboard sensors and assists in proportion to that estimate. However, its estimators are trained on data recorded with the target device, so deploying the controller on new hardware requires collecting new data. Existing open-source efforts release hardware, software, or controllers separately, so this data collection must still be repeated on the assembled device. Here, we present an open-source knee exoskeleton, built from off-the-shelf components and widely available fabrication tools, with a biological torque controller derived entirely from simulation. A muscle-driven imitation policy reproduces squatting motions from an open biomechanics dataset, and the resulting knee moments are distilled into an estimator that uses only onboard sensors, so no training data are recorded on the device. In seven able-bodied participants, assistance reduced peak biological knee extension moment by 21.5% and peak vastus medialis activation by 26.7% during squatting, relative to squatting without the exoskeleton (p<0.05). Our results demonstrate that a biological torque controller derived from simulation can transfer to physical hardware without device-specific data collection. Because the hardware and training pipeline are released together, research groups can train and compare knee exoskeleton controllers on shared hardware without additional data collection overhead.
Figures & tables
Fig. 1: Overview of the robotic knee exoskeleton design. (a) Sagittal-plane view of the unilateral device, with straps, hooks, and cuffs at the thigh, shank, and ankle. (b) A participant wearing the bilateral device, shown left to right in sagittal standing, sagittal squatting, and frontal standing. (c) The high-level control board built around the embedded computer (top) and the low-level control board built around the microcontroller (bottom).
Mechanical
Total mass, bilateral ∗ (kg)
3.5
Unilateral mass (kg)
1.3
Actuator mass (kg)
0.49
Rated / peak torque ∗∗ (Nm)
9 / 22
Range of motion (deg)
0–120
Compute/Electrical
TABLE I: Device Specifications
Fig. 2: Overview of the simulation-derived biological torque controller. An imitation policy is trained with reinforcement learning in a musculoskeletal simulation to reproduce reference squatting motion, using an imitation reward as its learning signal. Knee muscle moments from the policy rollouts are then distilled into a temporal convolutional network that estimates the knee moment from onboard sensor signals, trained by minimizing the error against the simulated moments. On the exoskeleton, the estimated knee moment is scaled and delayed to generate assistive torque.
Fig. 3: Experimental setup for validating the integrated exoskeleton and controller during squatting. A participant wearing the exoskeleton on the split-belt instrumented treadmill, whose embedded force plates measure ground reaction forces, with example placements of the electromyography sensors and reflective motion capture markers.
Fig. 4: Group-averaged knee kinematics, kinetics, and muscle activation during squatting. (a) Knee angle, (b) biological knee extension moment with commanded torque, (c) vastus medialis activation, and (d) medial gastrocnemius activation over the normalized squat cycle. Solid curves represent group means across participants, and shaded regions represent ± 1 standard deviation. Blue and brown indicate squatting without (No Exo) and with (Exo) the exoskeleton, respectively, and the dashed gray curve indicates the commanded torque with the exoskeleton. Muscle activation is expressed as a percentage of each participant’s maximum voluntary contraction (%MVC).
Fig. 5: Peak biological knee extension moment and muscle activation with and without the exoskeleton. (a) Peak biological knee extension moment, (b) peak vastus medialis activation, and (c) peak medial gastrocnemius activation without (No Exo) and with (Exo) the exoskeleton. Bars represent group means, and error bars represent ± 1 standard deviation across participants. Asterisks indicate a significant difference between conditions ( p<0.05 , paired t -test).
Device
Validation Subjects
Device Mass (kg)
Peak Torque (Nm)
Tasks
Reported Reduction
This Study
7 able-bodied
3.5
18
Squatting
26.7% vastus medialis activation
Nesler et al. [ 18 ]
1 able-bodied
1.6 a
30
Lifting and lowering
78.9% vastus medialis activation
Lee et al. [ 19 ]
10 able-bodied
4.4
15
Level walking, ramps, and stairs
6.2% metabolic cost c
Zhang et al. [ 34 ]
6 able-bodied
3.2
18
Level walking, ramp and stair ascent
4.3–12.0% metabolic cost
Boon et al. [ 16 ]
8 able-bodied
2.6 a
55
Stair ascent
26% vastus medialis activation
Chen et al. [ 13 ]
2 able-bodied
4.1
35
Kneeling
18.4% vastus medialis activation b,c
TABLE II: Comparison with Existing Knee Exoskeleton Platforms
Data-driven joint-moment predictors offer a scalable alternative to laboratory-based inverse-dynamics pipelines for biomechanics estimation and exoskeleton control. Meanwhile, physics-based reinforcement learning (RL) enables simulation-trained controllers to learn dynamics-aware assistance strategies without extensive human experimentation. However, quantitative verification of simulation-trained exoskeleton torque predictors, and their impact on human joint power injection, remains limited. This paper presents (1) an RL framework to learn exoskeleton assistance policies that reduce biological joint moments, and (2) a validation pipeline that verifies the trained control networks using an open-source gait dataset through inference and comparison with biological joint moments. Simulation-trained multilayer perceptron (MLP) controllers are developed for level-ground and ramp walking, mapping short-horizon histories of bilateral hip and knee kinematics to normalized assistance torques. Results show that predicted assistance preserves task-intensity trends across speeds and inclines. Agreement is particularly strong at the hip, with cross-correlation coefficients reaching 0.94 at 1.8 m/s and 0.98 during 5° decline walking, demonstrating near-matched temporal structure. Discrepancies increase at higher speeds and steeper inclines, especially at the knee, and are more pronounced in joint power comparisons. Delay tuning biases assistance toward greater positive power injection; modest timing shifts increase positive power and improve agreement in specific gait intervals. Together, these results establish a quantitative validation framework for simulation-trained exoskeleton controllers, demonstrate strong sim-to-data consistency at the torque level, and highlight both the promise and the remaining challenges for sim-to-real transfer.
Zihang You, Xianlian Zhou
Department of Biomedical Engineering, New Jersey Institute of Technology, Newark, NJ 07102, USA
Objective: Data-driven models that estimate physiological states, particularly biological joint moments, are widely used in exoskeleton control. However, these estimators are often coupled to device-specific sensor configurations, limiting controller transfer and the use of open-source biomechanics datasets. Methods: We proposed joint kinematics as an intermediate representation that decouples hardware-specific sensing from downstream biological joint moment estimation. A joint-moment estimator using joint angles and angular velocities was trained exclusively on open-source biomechanics data and evaluated using kinematics from a hip exoskeleton, knee exoskeleton, and inertial measurement unit (IMU) sensor suite during level-ground, ramp-ascent, and ramp-descent walking. Results: The estimator achieved an root mean square error (RMSE) of 0.17 Nm/kg and coefficient of determination (R2) of 0.79 using hip exoskeleton kinematics, 0.19 Nm/kg and 0.62 using knee exoskeleton kinematics, and 0.15 Nm/kg and 0.85 using kinematics derived from IMUs across bilateral hip, knee, and ankle joints. Conclusion: Joint kinematics enabled an estimator trained only on open-source data to operate across distinct wearable platforms. Significance: This framework may reduce target-device data collection and support transferable biological joint moment estimation for exoskeleton control and wearable biomechanics.
Developing exoskeleton controllers that generalize across diverse locomotor conditions typically requires extensive motion-capture data and biomechanical labeling, limiting scalability beyond instrumented laboratory settings. Here, we present a physics-based neuromusculoskeletal learning framework that trains a hip-exoskeleton control policy entirely in simulation, without motion-capture demonstrations, and deploys it on hardware via policy distillation. A reinforcement learning teacher policy is trained using a muscle-synergy action prior over a wide range of walking speeds and slopes through a two-stage curriculum, enabling direct comparison between assisted and no-exoskeleton conditions. In simulation, exoskeleton assistance reduces mean muscle activation by up to 3.4% and mean positive joint power by up to 7.0% on level ground and ramp ascent, with benefits increasing systematically with walking speed. On hardware, the assistance profiles learned in simulation are preserved across matched speed-slope conditions (r: 0.82, RMSE: 0.03 Nm/kg), providing quantitative evidence of sim-to-real transfer without additional hardware tuning. These results demonstrate that physics-based neuromusculoskeletal simulation can serve as a practical and scalable foundation for exoskeleton controller development, substantially reducing experimental burden during the design phase.
Ilseung Park, Changseob Song, Inseung Kang
Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, PA, 15213 USA.