Soft Robotic Exogloves for Dexterous Mobility -- Towards Personalized Rehabilitation

arXiv:2607.07968 · cs.RO · Submitted 2026-07-08 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: I'm Rosa, and with me are Dev and Taro, guest researcher.

Dev: Today's paper: "Soft Robotic Exogloves for Dexterous Mobility -- Towards Personalized Rehabilitation".

Rosa: The gist The authors present a personalized pneumatically actuated soft robotic exoglove designed to provide hand rehabilitation and assistance by tailoring actuator geometry to individual hand anatomy.

Dev: First, who's behind it and why it matters.

Paper summary: Rosa: So we're looking at this paper called "Soft Robotic Exogloves for Dexterous Mobility -- Towards Personalized Rehabilitation". Essentially, they’re saying that fitting these soft robotic gloves usually relies on standard measurements, which limits how well they actually work, especially when you need fine movement for things like delicate manipulation.

Dev: Right. The core idea here is to move away from those generic sizes and create a glove that's actually tailored to the individual hand. They design a personalized pneumatically actuated soft robotic exoglove using topological scans to fit the user's hand, and they fabricate it using silicone mold casting for the parts.

Taro: What’s really interesting is how they use that scan data to model things, like designing the dorsal base of the glove based on those contours. They also tailor the actuator geometry and joint placement directly based on measurements taken from Participant A's hand to align them with specific finger joints, like the metacarpophalangeal or MCP and proximal interphalangeal or PIP joints.

Rosa: And they don't just stop there; they run finite element analysis, which is a computer simulation, to test how those actuators bend and what kind of forces happen when there’s physical human-robot interaction between the actuator and a simplified biomechanical finger model.

Dev: That FEA part is important because it lets them analyze the physical human-robot interaction, or pHRI. They use this to understand the contact forces, which is crucial for making sure the movement feels right and doesn't cause unwanted pressure points on the user's hand.

Taro: It shows that anatomical personalization allows them to analyze those pHRI contact forces in a way that standard gloves just can't do. Plus, they found that even with non-ideal compression on the proximal phalanx, there was still sufficient joint mobilization.

Rosa: And finally, their pneumatic pressure control experiments showed that controlling the pressure lets them achieve accurate and targeted mobility for those MCP and PIP joints with some intrinsic stiffness built into the system.

Conclusion: Dev: So looking at this paper, "Soft Robotic Exogloves for Dexterous Mobility -- Towards Personalized Rehabilitation", what it really boils down to is moving from generic devices to highly customized ones. The authors claim that tailoring the glove structure based on individual hand anatomy through topological scans and then using FEA modeling to predict human-robot interaction helps them optimize the design for actual use.

Rosa: They are focusing on making sure that when you have this personalized exoglove, you get accurate mobility for the metacarpophalangeal and proximal interphalangeal joints because they controlled the pneumatic pressure to achieve that. It’s about achieving specific movements in a way that matches the actual finger structure.

Taro: The implication for people using these is that this personalization could lead to better assistance in dexterity tasks, and it lays groundwork for optimizing exoglove designs for helping with fine motor skills or neuromuscular rehabilitation of those small movements.

Dev: It suggests that if you can model the anatomical fit and simulate the contact forces accurately, you can design a system that provides much more effective assistance than what standardized measurements allow right now.

Rosa: So it's about making sure the hardware matches the human anatomy closely enough so that rehabilitation or assistance is actually helpful for complex hand movements.

Stevens Institute of Technology

cs.RO

Submitted: 2026-07-08

Updated: 2026-10-14

Comments: 8 pages, 14 figures. Accepted in IEEE Transactions on Medical Robotics and Bionics (T-MRB)

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 76/100

The gist: The gist The authors present a personalized pneumatically actuated soft robotic exoglove designed to provide hand rehabilitation and assistance by tailoring actuator geometry to individual hand

Key concepts

Topological Scans
These are 3D scans of a physical object, like a hand. The paper uses them to capture the exact shape and contours of a user's hand. This detailed digital model is crucial because it allows engineers to precisely tailor the design of the soft robotic glove, ensuring that the actuators fit correctly onto the specific anatomy rather than using generic sizes.
Finite Element Analysis (FEA)
FEA is a computer simulation technique used to test how a structure reacts to forces. In this study, FEA was used to predict how much bending and force would occur when the pneumatic actuators were activated against a simplified biomechanical finger model. This helps researchers understand the mechanical performance and safety of the glove before physical testing.
pHRI (Physical Human-Robot Interaction)
pHRI refers to the actual physical contact forces that occur when a human interacts with a robotic device, like wearing this exoglove. The study modeled these contact forces between an actuator and a finger using FEA. Understanding these specific interaction forces is vital for designing gloves that provide effective rehabilitation and assistance during real-world use.

Terminology

Summary

The gist The authors present a personalized pneumatically actuated soft robotic exoglove designed to provide hand rehabilitation and assistance by tailoring actuator geometry to individual hand anatomy.

Design and Personalization

**- Soft robotic exogloves can provide hand rehabilitation and assistance <ref:2607.07968#pg2> The glove was fit to a user’s hand with topological scans and fabricated with silicone mold casting <ref:2607.07968#pg4>. The glove was customized to Participant A <ref:2607.07968#pg3>. A 3D scan of Participant A’s left hand was taken using the Artec Eva 3D scanner <ref:2607.07968#pg3> Measurements of each finger were taken to personalize the size of the actuators <ref:2607.07968#pg2>. The contours of the hand scan were used to model the dorsal base <ref:2607.07968#pg4>. The actuator geometry and joint placement are designed based on hand measurements to align bending regions of the actuator with the metacarpophalangeal (MCP) and proximal interphalangeal (PIP) joints <ref:2607.07968#pg2>. The number of air chambers on each actuator joint is based on the size of the corresponding finger joint <ref:2607.07968#pg4>. The length of the proximal phalanx determines the location of the MCP and PIP actuator joints to be centered with their corresponding finger joints <ref:2607.07968#pg4>. The actuator body does not extend to the DIP joint <ref:2607.07968#pg4>. To measure the bending angle, a flex sensor is embedded in the base layer of each actuator <ref:2607.07968#pg4>. A custom rigid dorsal base to secure the actuators was derived from the 3D hand scan to match the contours of the hand for consistent actuator placement <ref:2607.07968#pg4>. The exoglove base design also has features for integrating the electrical and pneumatic components with the human hand <ref:2607.07968#pg4>. The exoglove was 3D printed out of standard polylactic acid (PLA) plastic using a Bambu Lab X1 Carbon FDM printer with 15% infill <ref:2607.07968#pg5>. Molds were used to fabricate the actuators <ref:2607.07968#pg5>. The silicone used to fabricate the actuators was Sorta Clear 40 (Smooth-On, Inc.), a two-part platinum cure silicone rubber with a Shore hardness of 40A and a tensile strength of 800 psi <ref:2607.07968#pg6>. The air inlets are fabricated and cured first, as they will be affixed to the main layer mold before pouring the main layer <ref:2607.07968#pg5>. Silicone is poured into the bottom mold to the height the Velcro blocks before they are inserted to prevent air entrapment <ref:2607.07968#pg5>. After inserting the blocks, the remaining silicone is poured into the mold, and the lid is placed on top <ref:2607.07968#pg5>. The exoglove base was designed from the 3D scanned topology of the dorsal side of the hand <ref:2607.07968#pg4>. The exoglove base design also has features for integrating the electrical and pneumatic components with the human hand <ref:2607.07968#pg4>. Clamps are added to the base such that the actuators are anchored and aligned to the fingers <ref:2607.07968#pg4>. Clips are placed to secure and organize the tubing and wiring <ref:2607.07968#pg4>. Loops for Velcro allow straps to secure the palm of the hand to the base <ref:2607.07968#pg4>. The actuator features two joints; one to bend the PIP joint (green) and the other to bend the MCP joint (blue) <ref:2607.07968#pg4>. The number of chambers of each joint of the actuator was based on the size of the corresponding joint of the finger <ref:2607.07968#pg4>. To personalize the size of the actuator to the hand, measurements of the proximal phalanx were taken to determine the distance between the two joints <ref:2607.07968#pg4>. The material model for the actuator is Ecoflex 00-50 material, which has a shore hardness similar to human flesh <ref:2607.07968#pg5>. The biomechanical finger was modeled with Ecoflex 00-50 material, which has a shore hardness similar to human flesh <ref:2607.07968#pg5>. Frictionless contacts were added between the bottom face of the actuator and the top faces of the finger <ref:2607.07968#pg5>. The internal pressure ramp of 0 to 40 kPa (5.8 psi) was applied to the actuator, and the results at the maximum pressure are presented in Figs. 12(a,c) <ref:2607.07968#pg5>. A simplified biomechanical finger model was created based on Participant A’s dimensions <ref:2607.07968#pg5>. The actuator was positioned in the same place that it would be when worn by the participant <ref:2607.07968#pg5>. The material model for the actuator is the same as in the previous free-bending simulation <ref:2607.07968#pg5>. Frictionless contacts were added between the bottom face of the actuator and the top faces of the finger <ref:2607.07968#pg5>. The biomechanical finger was modeled with Ecoflex 00-50 material, which has a shore hardness similar to human flesh <ref:2607.07968#pg5>. The internal pressure ramp of 0 to 40 kPa (5.8 psi) was applied to the actuator, and the results at the maximum pressure are presented in Figs. 12(a,c) <ref:2607.07968#pg5>. A simplified biomechanical finger model was created based on Participant A’s dimensions <ref:2607.07968#pg5>. The actuator was positioned in the same place that it would be when worn by the participant <ref:2607.07968#pg5>. The material model for the actuator is the same as in the previous free-bending simulation <ref:2607.07968#pg5>. Frictionless contacts were added between the bottom face of the actuator and the top faces of the finger <ref:2607.07968#pg5>. The biomechanical finger was modeled with Ecoflex 00-50 material, which has a shore hardness similar to human flesh <ref:2607.07968#pg5>. The internal pressure ramp of 0 to 40 kPa (5.8 psi) was applied to the actuator, and the results at the maximum pressure are presented in Figs. 12(a,c) <ref:2607.07968#pg5>. A simplified biomechanical finger model was created based on Participant A’s dimensions <ref:2607.07968#pg5>. The actuator was positioned in the same place that it would be when worn by the participant <ref:2607.07968#pg5>. The material model for the actuator is the same as in the previous free-bending simulation <ref:2607.07968#pg5>. Frictionless contacts were added between the bottom face of the actuator and the top faces of the finger <ref:2607.07968#pg5>. The biomechanical finger was modeled with Ecoflex 00-50 material, which has a shore hardness similar to human flesh <ref:2607.07968#pg5>. The internal pressure ramp of 0 to 40 kPa (5.8 psi) was applied to the actuator, and the results at the maximum pressure are presented in Figs. 12(a,c) <ref:2607.07968#pg5>. A simplified biomechanical finger model was created based on Participant A’s dimensions <ref:2607.07968#pg5>. The actuator was positioned in the same place that it would be when worn by the participant <ref:2607.07968#pg5>. The material model for the actuator is the same as in the previous free-bending simulation <ref:2607.07968#pg5>. Frictionless contacts were added between the bottom face of the actuator and the top faces of the finger <ref:2607.07968#pg5>. The biomechanical finger was modeled with Ecoflex 00-50 material, which has a shore hardness similar to human flesh <ref:2607.

Improvements for AI systems

  1. No strain-limiting layer requirement for alignment: From these tests, it can be concluded that the strain-limiting layer is not required to achieve bending and hinders the performance of the actuator for this application. This allows actuators to maintain proper alignment with human joints, improving kinematic fidelity during actuation.

  2. Anatomically personalized joint placement: The actuator geometry and joint placement are designed based on hand measurements to align bending regions of the actuator with the metacarpophalangeal (MCP) and proximal interphalangeal (PIP) joints. This precise tailoring ensures that actuator joints shift away from the finger joints less, leading to better functional assistance.

  3. Pressure control as a force/position control benchmark: Pressure control was chosen as a benchmark for future position or force control, since the measurement of air pressure is the most fundamental physical element to sense after air flow. This enables the development of more robust and safety-oriented robotic controllers for soft devices by leveraging pressure readings.

  4. pHRI analysis for force distribution insights: Fig. 12c shows a top view of the von Mises stress plot of just the human finger model, with the soft actuator hidden, focusing on this lower stress distribution range. This analysis provides data to inform design optimization to minimize unwanted axial compressive forces on the proximal phalanx.

  5. Closed-loop PID control for dynamic interaction: The use of a closed-loop PID controller was implemented to regulate the internal pressure of each soft pneumatic actuator allows the system to reject external forces while providing compliance, enabling safe and comfortable human–robot interaction in dynamic scenarios.

Sources

Related papers