A Task-Agnostic Control Strategy for Dynamic Assistance with Pneumatically Actuated Soft Exosuits

Anoush Sepehri; Zachary Huang; Raymond de Callafon; Michael T. Tolley; Tania K. Morimoto

cs.ROAug 18, 2026

A Task-Agnostic Control Strategy for Dynamic Assistance with Pneumatically Actuated Soft Exosuits

Authors: Anoush Sepehri, Zachary Huang, Raymond de Callafon, Michael T. Tolley, Tania K. Morimoto

Organizations: Department of Mechanical and Aerospace Engineering, University of California San Diego, San Diego, CA 92093, USA

Abstract

Pneumatic artificial muscles have provided new opportunities to develop upper-extremity soft exosuits for reha- bilitation, augmentation, and assisted daily living. However, the complex dynamics and limited bandwidth of these actuators has made providing responsive assistance based on user intention a longstanding challenge. In this work, we present an inverse-plant control strategy for pneumatically actuated soft exosuits that only relies on kinematic sensing for task-agnostic and dynamic assistance during daily living. We model the human-robot system using a Hammerstein dynamic model, consisting of a Preisach hysteresis model and a linear time-invariant filter, to capture the static and dynamic behavior of the system. We personalize our model to each user using 140 s of data and approximate an inverse to integrate into our control loop. When evaluated on a test rig that emulated a soft assistive exosuit for the wrist, our controller reduced the interaction torque by up to 73% and the activation of key flexor and extensor muscles by up to 47% relative to the condition with no assistance for speeds ranging from 8°/s to 120°/s. Overall, this work presents a control strategy that can provide task-agnostic, dynamic assistance with pneumatically actuated soft exosuits without the need for physiological or force sensors to interpret user intention.

Figures & tables

Fig. 1: An idealized block diagram of the human-robot system (P), the proposed, positive feedback controller (C), and the human’s internal motor controller (H). The controller has access to the kinematic state of the robot

(\theta)

and generates assistive torques

(\tau_{r}

) to augment the user’s effort

(\tau_{h}

)

Fig. 2: Simulation results to evaluate our proposed control strategy. (a) We expanded on the block diagram in Fig. 1 to include muscle dynamics and coupling dynamics between the human and the robot. (b) While existing static control techniques can provide effective assistance (i.e., a reduction in the effort,

\tau_{h}

) in quasi-static scenarios, they fail to provide meaningful assistance at higher frequencies necessary for dynamic movements.

Fig. 3: Overview of the test rig for flexion and extension assistance for the wrist. (a) The test rig consisted of two pouch motors in parallel with the flexor and extensor muscles of the wrist and linear springs to simulate elevated joint stiffness after a neurological injury. The wrist support and anchors for the proximal end of the pouch motors can be adjusted to account for anatomical differences between users. (b) The experimental test setup consisted of the test rig and a visual guide for the marker following tasks. (c) The passive stiffness of the test rig with $(k_{spring})$ and without $(k_{exo})$ the linear springs to simulate elevated joint stiffness. The stiffness without the springs was due to the soft coupling to the pouch motors. (d) Force-strain characterization results for the small and large pouch motors used in this work.

Fig. 4: The inverse-plant control strategy developed in this work (a) The (i) human-robot system represented as a Hammerstein dynamic model with the (ii) robot controller. (b) Experimental characterization results for a representative user to identify the parameters in the Hammerstein dynamic model. Left: The hysteresis properties of the human-robot system that maps the reference pressure (

u_{r}

) to the joint position (

v(r)

) for quasi-static pressure trajectories. Right: The empirical transfer function, based on spectral analysis, to characterize the pressurized air and soft coupling dynamics for the small and large pouch motors used in this work. (c) Representative results of the proposed inverse dynamic filter based on the identified plant dynamics from (b). The inverse dynamic filter was cascaded with a parameter-varying filter

(H(z,\rho))

that attenuated high frequency dynamics based on the scheduling variable

\rho

Fig. 5: Summary of the performance of different model structures when used as an inverse-plant feedforward controller. (a) Normalized root mean squared tracking error (NRMSE) and normalized lag between the reference and actual position for different sinusoidal trajectories when using the small pouch geometry. The dynamic models (D1-D3) significantly outperformed the static models (S1-S3) for the 1.0 Hz and 2.0 Hz trajectories. (b) Performance of the static (S3) and dynamic (D3) model for different pouch geometries. The dynamic model accounted for the changes in air flow dynamics between the small and large pouches to achieve high tracking accuracy regardless of the pouch geometry. Differences between groups that were statistically significant were marked by an asterisk.

Fig. 6: Effects of the parameter varying filter,

H(z,\rho)

, on the mean absolute acceleration for the 8°/s (0.1 Hz) trajectory. (a) Representative results for the dynamic controller from a single trial. Without

H(z,\rho)

, the inverse dynamic controller amplified high frequency noise which led to non-smooth motions. (b) Mean absolute acceleration results for all users and test conditions. Statistically significant differences between controllers were marked with the black asterisks while statistically significant differences between the controllers and the no control condition were marked with the colored asterisks.

Fig. 7: Representative results from a user for the 80°/s (1.0 Hz) trajectory. (a) Participants tried to follow a reference trajectory that was presented on the screen. We recorded their position, the calculated reference pressure based on the selected controller, and the interaction torque between the user and the robot. (b) We also monitored the activation of the flexor carpi ulnaris (FCU), flexor carpi radialis (FCR), extensor carpi ulnaris (ECU), and extensor carpi radialis (ECR) muscles to evaluate the physiological effects of the different controllers.

Fig. 8: Interaction torque and muscle activation results for trajectories with reference speeds ranging from 8°/s (0.1 Hz) to 120°/s (1.5 Hz). (a) Change in the mean absolute interaction torque for the different control conditions. The dynamic controller outperformed the static controller for the 80°/s (1.0 Hz) and 120°/s (1.5 Hz) trajectories. (b) Change in the mean muscle activation for the flexor and extensor muscles in the wrist. The dynamic controller once again provided more effective assistance for the 80°/s (1.0 Hz) and 120°/s (1.5 Hz) trajectories. Statistically significant differences between controllers were marked with the black asterisks while statistically significant differences between the controllers and the no control case were marked with the colored asterisks.

Fig. 9: Interaction torque and muscle activation results for the target reaching task. (a) Representative results from a user for the target reaching task when testing the dynamic controller. (b) Change in the mean absolute interaction torque between the user and the test rig for the static and dynamic controller conditions. (c) Change in the mean muscle activation for the flexor and extensor muscles in the wrist. We observed similar performance between the static and dynamic controllers relative to the no control condition. Statistically significant differences between the controllers and the no control case were marked with the colored asterisks.

Assistive Robotics

Assistive robotics develops robots and wearable devices that support mobility, daily activities, rehabilitation, and care.

See all 199 papers

Fabric Pneumatic Artificial Muscle-Based Head-Neck Exosuit: Design, Modeling, and Evaluation
Mar 13, 2026Katalin Schäffer, Ian Bales, Haohan Zhang +1Soft Pneumatic Actuators Assistive Robotics
Department of Aerospace and Mechanical Engineering, University of Notre Dame, Notre Dame, IN 46556, USA. · Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, 1083 Budapest, Hungary. · Department of Mechanical Engineering and Robotics Center, University of Utah, Salt Lake City, UT 84112 USA.
Task-space model-based control of pneumatic soft actuators
Aug 27, 2026Nithin S. Kumar, Joshua Gaston, D. Caleb Rucker +1Robotic Control Soft Robotics
Department of Mechanical Engineering, Vanderbilt University, Nashville, Tennessee 37235 · Department of Mechanical, Aerospace, and Biomedical Engineering, The University of Tennessee, Knoxville, TN 37996
Koopman Model Predictive Control of An Origami-Inspired Soft Exoskeleton for Knee Rehabilitation
Oct 13, 2025Junxiang Wang, Han Zhang, Zehao Wang +3Robotic Control Model Predictive Control
School of Automation and Intelligent Sensing, Institute of Medical Robotics, Shanghai Jiao Tong University, and Key Laboratory of System Control and Information Processing, Ministry of Education of China, Shanghai 200240, China · Seventh Affiliated Hospital, SunYat-sen University, Shenzhen 510275, China

Mar 13, 2026J/K move · Enter open · S save

Fabric Pneumatic Artificial Muscle-Based Head-Neck Exosuit: Design, Modeling, and Evaluation

Katalin Schäffer, Ian Bales, Haohan Zhang, Margaret McGuinness

Department of Aerospace and Mechanical Engineering, University of Notre Dame, Notre Dame, IN 46556, USA. · Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, 1083 Budapest, Hungary. · Department of Mechanical Engineering and Robotics Center, University of Utah, Salt Lake City, UT 84112 USA.

Soft Pneumatic Actuators Assistive Robotics Soft Robotics

Open paper AskPDF arXiv

Wearable exosuits assist human movement in tasks ranging from rehabilitation to daily activities; specifically, head-neck support is necessary for patients with certain neurological disorders. Rigid-link exoskeletons have shown to enable head-neck mobility compared to static braces, but their bulkiness and restrictive structure inspire designs using "soft" actuation methods. In this paper, we propose a fabric pneumatic artificial muscle-based exosuit design for head-neck support. We describe the design of our prototype and physics-based model, enabling us to derive actuator pressures required to compensate for gravitational load. Our modeled range of motion and workspace analysis indicate that the limited actuator lengths impose slight limitations (83% workspace coverage), and gravity compensation imposes a more significant limitation (43% workspace coverage). We introduce compression force along the neck as a novel, potentially comfort-related metric. We further apply our model to compare the torque output of various actuator placement configurations, allowing us to select a design with stability in lateral deviation and high axial rotation torques. The model correctly predicts trends in measured data where wrapping the actuators around the neck is not a significant factor. Our test dummy and human user demonstration confirm that the exosuit can provide functional head support and trajectory tracking, underscoring the potential of artificial muscle-based soft actuation for head-neck mobility assistance.

Abstract

Assistive Robotics

Explore similar work

Fabric Pneumatic Artificial Muscle-Based Head-Neck Exosuit: Design, Modeling, and Evaluation

Task-space model-based control of pneumatic soft actuators

Koopman Model Predictive Control of An Origami-Inspired Soft Exoskeleton for Knee Rehabilitation