Assistive Robotics Laboratory · Research program
Our research addresses how robotic systems generate assistance, how that assistance can be customized, and how its effects on human movement are evaluated. Projects span lower-limb exoskeletons, length-actuated prostheses, and compliant robotic mechanisms.
Our approach
Conventional control methodologies for lower-limb exoskeletons often rely on pre-defined reference kinematics or task-specific torque profiles. These approaches can constrain voluntary motion and require parameter adjustment across users and locomotor tasks.
We investigate reference kinematics-free control strategies based on energy shaping, centroidal momentum regulation, and control barrier functions. Related research addresses length-actuated prostheses, continuum robotic systems, and passivity-based bipedal locomotion. Mathematical analysis, simulation, and human-subject experiments are used to characterize controller performance and human–robot interaction.
Develop task-agnostic control strategies and customize robotic assistance through human-in-the-loop optimization.
Develop and prototype user-friendly wearable devices and compliant mechanisms for physical interaction with human users.
Conduct biomechanics studies to characterize human movement and evaluate assistance through muscle activity, joint mechanics, and stability-related outcomes.
01 / Control and Customization
Control formulations and optimization frameworks for assistance that accommodates user-selected movement.
Control formulation · Energy and momentum
Task-specific reference trajectories can restrict the adaptability of lower-limb exoskeletons to continuously varying human activities. We investigate control paradigms that regulate human body energetics and centroidal momentum without prescribing reference joint kinematics.
Kinematic control approaches depend on pre-defined reference trajectories that can be difficult to adjust across locomotor tasks and human subjects. For individuals with partial or full voluntary control of their lower limbs, such references may overly constrain self-selected motion.
Energy shaping modifies the closed-loop mechanical energy of the coupled human–exoskeleton system. Potential energy shaping can provide body-weight support, whereas kinetic energy shaping modifies effective mass and inertia. Highly backdrivable actuation facilitates voluntary movement while enabling the commanded assistive torques.
Our centroidal momentum shaping paradigm regulates the sum of limb momenta projected onto the human center of mass. A virtual reference model is constructed from each user’s real-time gait with scaled limb inertial parameters. A nonlinear disturbance observer estimates voluntary human joint torques, and these estimates are incorporated into the control design.
Our 2026 IEEE Robotics and Automation Letters study reports preliminary experiments with six non-disabled participants wearing a bilateral hip exoskeleton during treadmill walking, ground–stair transitions, and sit-to-stand tasks. Ensemble-averaged muscle efforts decreased by 22% with assistance. These findings are specific to the preliminary study conditions.
IEEE Robotics and Automation Letters · 2026
Preliminary validation of centroidal momentum shaping on a hip exoskeleton.
Watch on YouTube ↗IEEE ICRA · 2025
Conference video presenting preliminary validation of task-agnostic assistance on a bilateral hip exoskeleton.
Watch on YouTube ↗Control customization · Human-in-the-loop optimization
The relationship between control parameters and human effort varies across individuals and locomotor activities. We combine task-agnostic assistive strategies with sample-efficient optimization to customize lower-limb exoskeleton assistance.
Although energy-shaping strategies can provide task-agnostic assistance, their parameters influence the resulting human joint torques and energetic demands. A systematic customization framework is needed to identify appropriate parameters across users and tasks.
We investigate derivative-free, sample-efficient optimization in conjunction with energy-shaping control. The framework searches for assistive parameters using human performance-based objectives while retaining a task-agnostic controller structure.
The customization loop
Simulation studies of the learning framework demonstrate reductions in modeled human joint torque and metabolic cost. Related work investigates a two-layer human-in-the-loop optimization structure and physics-informed neural networks for solving the energy-shaping matching condition. Simulation findings are distinguished from experimental validation.
Stability augmentation · Perturbed locomotion
The vast majority of lower-limb exoskeleton controllers are developed and validated under steady-state locomotor conditions. We investigate proactive stability augmentation under leaning and slip perturbations while preserving voluntary human motion.
Control paradigms validated under nominal walking may not generalize to unstable conditions or external perturbations. In addition, task-specific recovery trajectories can constrain the user’s voluntary responses.
The proposed paradigm defines a center of mass-based stability indicator and enforces its dynamics within a user-specific safe range through control barrier functions. Corrective assistance is generated as the system state approaches the safety-set boundary. A nonlinear disturbance observer estimates a lumped input comprising voluntary human joint torque and external perturbations; this estimate is incorporated into the control design.
The research question
Proactive stability assistance without pre-defined reference kinematics.
The controller accounts for user-specific stability constraints, voluntary joint torques, and external perturbations within a common control formulation.
The final Humanoids 2026 manuscript reports preliminary experiments with two non-disabled participants wearing a bilateral hip exoskeleton. Ensemble-averaged muscle efforts decreased by 41% during leaning and 12% during slip perturbations, with lower peak-to-peak whole-body angular momentum and faster recovery. These preliminary results do not establish efficacy outside the tested participants and conditions.
IEEE-RAS Humanoids · 2026
Preliminary experimental validation of a control barrier function-based stability augmentation paradigm on a bilateral hip exoskeleton.
Watch on YouTube ↗Conference paper (PDF) →IEEE Control Systems Letters · Simulation
Simulation accompanying the balance-control case study by Miao Yu, Qin Lin, and Ge Lv.
Watch on YouTube ↗02 / Design
Device architectures, compliant mechanisms, and prototypes evaluated in the laboratory’s research.
Device development · Prosthetic systems
We investigate dynamic leg-length actuation as an alternative prosthetic design methodology. The objective is to coordinate stance support and swing-phase clearance through active shortening and lengthening of a transfemoral prosthesis.
Transfemoral prosthesis users may experience challenges in stance stability, push-off, and swing-phase toe clearance. These requirements motivate prosthetic mechanisms that actively modify leg length during ambulation.
The proposed device employs a prismatic mechanism to modify leg length during the gait cycle. Dynamic leg-length modeling informs the desired actuation profile, and experimental evaluation incorporates user preferences and gait biomechanics.
Experiments with participants with amputation investigate gait symmetry, stability, and mobility with the length-actuated device. The design and preliminary evaluation are reported in IEEE Access (2024), with a subsequent case study in Prosthetics and Orthotics International (2026).
IEEE Access · Design and preliminary evaluation
Research video accompanying the length-actuated prosthesis study by Parr, DesJardins, Hippensteal, Harvey, and Lv.
Watch on YouTube ↗Compliant mechanisms · Continuum robotics
We investigate continuum and hybrid rigid–compliant robotic systems for physical interaction with human users and unstructured environments. Structural compliance provides a means of accommodating contact and geometric variation during assistance and manipulation.
Continuum robots achieve distributed bending through compliant structures. We investigate their use as auxiliary limbs and wearable appendages for daily assistance, including the preliminary design and evaluation of a continuum robot-based fall-mitigation system.
Related work explores a continuum-legged biped with a kangaroo-inspired tail for tripedal locomotion, extending compliant mechanisms to new robot architectures.
In collaboration with Prof. Ian Walker’s group, we investigate an avian-inspired gripper that integrates actuated rigid structures with compliant continuum digits. The hybrid architecture combines load-bearing capability with adaptability during grasping. Related work develops a teleoperation framework for the compliant gripper.
IEEE CASE · 2025
Demonstration of teleoperation of a compliant gripper with continuum digits.
Watch on YouTube ↗Watch supplementary video on MDPI ↗
Robotics · 2024 · Supplementary Video S1
Supplementary video accompanying the Robotics paper, showing experiments with the avian-inspired hybrid rigid-continuum gripper.
Supplementary Video S1 on MDPI ↗Control foundations
Passivity-based control provides a foundation for reasoning about energy, stability, and interaction in legged systems.
Our bipedal locomotion work introduces a generalized energy storage function that accounts for energy stored and dissipated by an inner-loop controller. An outer-loop passivity-based controller is designed to enlarge the basin of attraction, improve robustness to ground slope, and accelerate convergence to a stable walking cycle.
Explore the foundational work →Research support
Awarded research support includes National Science Foundation projects and a faculty development award through the NIH-supported SC-TRIMH program.
National Science Foundation
Foundational Research in Robotics · CAREER
This project investigates task-agnostic assistive strategies and a human-in-the-loop framework for rapidly customizing lower-limb exoskeleton assistance across users and locomotor tasks.
NSF award record →National Science Foundation
Human-Centered Computing · Collaborative Research: HCC: Small
Collaborative research on robotic systems for domestic activity. The subaward was transferred from the previous principal investigator, Ian Walker, University of Wyoming.
National Institutes of Health · SC-TRIMH
Phase I · Center of Biomedical Research Excellence (COBRE)
Faculty development support through SC-TRIMH under Phase I of the NIH Center of Biomedical Research Excellence program.
Clemson faculty profile →Any opinions, findings, conclusions, or recommendations expressed in NSF-supported work are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Explore further
Our research integrates control theory, robotic hardware development, and experimental evaluation. Related publications and research personnel are available below.
