Soft Fabric Omnidirectional Bending Actuator
Designed and characterized a lightweight, fabric-based soft actuator platform that bends in any direction, can be built with accessible fabrication tools, and supports programmable stiffness changes through coordinated chamber pressurization. The project focused on making a high-performance soft actuator easier to reproduce and more practical for future wearable and continuum-robot applications.
Key Results:
Developed a low-cost, easy-to-fabricate soft continuum actuator framework achieving 90.8° omnidirectional bending, 10.99 N payload, 862 kPa pressure tolerance, and a 39.53 N/kg force-to-weight ratio.
Top and front views of the actuator demonstrating multi-directional bending and full omnidirectional motion capability.
Problem
Soft robots are well suited for safe, adaptive interaction, but fewer actuator designs combine omnidirectional bending, low mass, strong force output, and repeatable fabrication in one system. Fabric-based actuators are especially promising because they are lightweight, compliant, low-profile when unpressurized, and capable of high force-to-weight performance. However, despite these advantages, multi-degree-of-freedom fabric actuators have remained relatively underexplored. The challenge was to create a fabric soft actuator that was easy to build, mechanically capable, and versatile enough for practical robotic applications.
Solution
I developed the Omnidirectional Bending Actuator (OBA), a three-chamber pneumatic soft actuator built from knit fabrics with 3D-printed end caps. By arranging the three chambers 120° apart and exploiting fabric anisotropy, the actuator can bend across 360° rather than only in one plane. Relative to the previous knit-fabric state-of-the-art benchmark, the design reduced total actuator mass from 400 g to 278 g and increased force-to-weight ratio from 24.15 to 39.53 N/kg, achieving a 31% mass reduction and 64% improvement in force-to-weight ratio. The same architecture also supported antagonistic stiffening through coordinated chamber pressurization, reaching up to 3.22× baseline stiffness.
Exploded view of the fabric actuator architecture, showing the anisotropic fabric body, internal bladder, stitched chamber structure, pneumatic connections, and 3D-printed end caps.
Design and fabrication
The actuator used three pneumatic chambers arranged around its circumference. Each chamber was built from two fabric layers with different mechanical properties: an extensible layer that expanded more under pressure, and a strain-limiting layer that expanded less. When the chamber was pressurized, this unequal expansion created a length difference, causing it to bend toward the strain-limiting side. By placing three of these chambers around the actuator body, the system could bend in different directions depending on which chamber or combination of chambers was pressurized.
A major engineering challenge was achieving repeatable bending. Early versions buckled unpredictably, causing out-of-plane motion and cycle-to-cycle inconsistency. I solved this by widening the inextensible layers so adjacent chambers could be stitched together along their lengths, then adding a pre-buckling band constraint around the actuator midsection. That combination forced buckling to occur at the same location each cycle and made the motion far more repeatable.
Repeatable bending mechanism showing the stitched chamber layout and pre-buckling band constraint used to suppress out-of-plane motion and enforce consistent bending behavior.
I also accounted for material conditioning as part of the fabrication framework. Because the bladder and textile layers softened after cycling, newly built actuators initially had different motion than conditioned ones. After about 60 fatigue cycles, bending behavior became much more stable, and by 120 cycles the bending angle had increased from 54.2° to 90.8° and largely stabilized.
The fabrication process was intentionally streamlined to make the actuator low-cost, accessible, and easy to reproduce. Each chamber was assembled from off-the-shelf materials using a sewing machine, while 3D-printed end caps provided sealing and mechanical integration. By reducing the required toolchain to a sewing machine and 3D printer, the design avoided reliance on shop-scale equipment such as heat presses, CNC machines, or laser cutters. The finished actuator measured 229 mm in length and 68 mm in diameter.
Actuator response before and after fatigue conditioning, showing improved motion capability after cycling.
To fully characterize the hardware, I built two dedicated benchtop test platforms. The first used an EM sensor attached to the free end of the actuator to map workspace and orientation under different pressure combinations. The second used a 6-DoF load cell on a linear rail so the actuator could extend slightly during pressurization, allowing bending torque to be measured more cleanly without contamination from extension induced moments. These rigs let me quantify workspace, torque output, and repeatability of the actuator quantitatively.
Custom benchtop test platforms used to quantify actuator workspace, orientation, bending torque, and repeatability under controlled pressure inputs.
Finally, I demonstrated that the actuator could also compensate for out-of-plane deformation under load by actively redistributing chamber pressures. Under a 100 g payload, the tip initially deflected by about 26 mm, but pressure retuning brought the actuator back close to its original pose within ±1.9 mm. This demonstrated that the actuator could actively compensate for load-induced deformation and adapt its shape to changing external conditions.
Load compensation experiment showing how pressure retuning restored tip position after payload-induced deflection.
Related Publication
Design and characterization of soft fabric omnidirectional bending actuators.
Lee, K., Bayarsaikhan, K., Aguilar, G., Realmuto, J., & Sheng, J. (2024).
Actuators. DOI: 10.3390/act13030112.