cs.ROMar 16, 2026
SaveMulti-Mode Pneumatic Artificial Muscles Driven by Hybrid Positive-Negative Pressure
Organizations: Department of Mechanical Engineering at Tsinghua University, Beijing, 100084, China. · Beijing Key Laboratory of Transformative High-end Manufacturing Equipment and Technology, Tsinghua University, Beijing 100084, China.
Abstract
Artificial muscles embody human aspirations for engineering lifelike robotic movements. This paper introduces an architecture for Inflatable Fluid-Driven Origami-Inspired Artificial Muscles (IN-FOAMs). A typical IN-FOAM consists of an inflatable skeleton enclosed within an outer skin, which can be driven using a combination of positive and negative pressures (e.g., compressed air and vacuum). IN-FOAMs are manufactured using low-cost heat-sealable sheet materials through heat-pressing and heat-sealing processes. Thus, they can be ultra-thin when not actuated, making them flexible, lightweight, and portable. The skeleton patterns are programmable, enabling a variety of motions, including contracting, bending, twisting, and rotating, based on specific skeleton designs. We conducted comprehensive experimental, theoretical, and numerical studies to investigate IN-FOAM's basic mechanical behavior and properties. The results show that IN-FOAM's output force and contraction can be tuned through multiple operation modes with the applied hybrid positive-negative pressure. Additionally, we propose multilayer skeleton structures to enhance the contraction ratio further, and we demonstrate a multi-channel skeleton approach that allows the integration of multiple motion modes into a single IN-FOAM. These findings indicate that IN-FOAMs hold great potential for future applications in flexible wearable devices and compact soft robotic systems.
Figures & tables
Fig. 1: Design and fabrication of IN-FOAMs. (a) A linear IN-FOAM made of TPU-coated nylon fabrics. (b) Materials used to fabricate the actuator. The sheet materials have been patterned through engraving. (c) Working principle. The actuator is flat when not actuated. Skeleton inflation induces tension in the sheets to generate contraction, and voids appear between skeleton columns. After applying a vacuum to the voids, the difference in pressure relative to the atmosphere induces tension in the skin and drives the actuator to contract further. (d) A pneumatic circuit design of IN-FOAMs. (e) Fabrication process. IN-FOAMs can be fabricated rapidly in four steps: (step 1) pattern the sheets, (step 2) stack and heat-press the sheets, (step 3) cut and heat-seal layers of the skeleton, (step 4) add pneumatic connectors and heat-seal the skin.
Fig. 2: Various motions performed by IN-FOAMs. The actuators’ skeletons are made of TPU-coated nylon fabrics (thickness: 0.09 mm), and the skins are made of transparent TPU film (thickness: 0.1 mm). Without actuation, the actuators are flat. (a) A linear IN-FOAM with a parallel-pouch skeleton (size: , in the initial state; weight: 6.26 g) contracts 17 mm driven by positive pressure and further contracts 7 mm under the negative-pressure actuation. (b) A parallel-pouch skeleton with an inflated beam at its right side enables the actuator (size: , in the initial state; weight: 4.31 g) to bend 47° leftwards driven by positive pressure and further bend 8° under the negative-pressure actuation. (c) With inflated beams to constrain contraction of the bottom skeleton, the actuator (size: , in the initial state; weight: 5.97 g) with a parallel-pouch skeleton is allowed to bend 39° upwards and further bend 23° under positive and negative pressure actuation, respectively. (d) By an anti-symmetric arrangement of oblique parallel pouches in the upper and bottom skeleton, the actuator (size: , in the initial state; weight: 6.65 g) twists 79° and further twists 13° under the positive and negative pressure actuation, respectively. (e) A flasher-origami-like skeleton makes the actuator (size: , in the initial state; weight: 10.34 g) rotate nearly 23° driven by positive pressure and further rotate 10° under the negative-pressure actuation. (f) By patterning the skeleton with L-shaped units, the actuator (size: , in the initial state; weight: 58.80 g) generates a length decrease of 36 mm and a width decrease of 41 mm under the positive pressure actuation; under the negative pressure actuation, the length and the width further decrease by 34 mm and 13 mm, respectively.
Fig. 3: Illustration of the modeling for the linear IN-FOAM. Cross-sectional Views of the FE simulation results are shown in Fig. A2 (a-d). (a) Model A: the skin between skeleton columns has no contact with the skeleton under the heat-pressed boundary (denoted by point A). (b) Model B: the skin contacts with the skeleton under point A as the linear IN-FOAM contracts further. (c) Model C: when the vacuum is high enough, the upper and bottom skin between skeleton columns come into contact on the basis of Model A. (d) Model D: when the vacuum is high enough, the upper and bottom skin between skeleton columns come into contact on the basis of Model B.
Fig. 4: Static property of the linear IN-FOAM with constant positive and negative pressures. (c-f) Experimental, theoretical, and simulation results of blocked force and force-contraction. Exp.: experiment; FE: finite element simulation. In the theoretical model, the elastic modulus is set to 400 MPa, which is roughly in accord with the material’s tensile test results. The factor (4 %) is estimated by measuring the change in a sample’s length before and after the heat-pressing process. (a) The tested sample (size: , in the initial state; weight: 24.6 g; skeleton length: 20 cm) at the maximum contraction ( 43 %, when kPa, kPa). (b) The simulated deformation at the maximum contraction, when kPa, kPa. The contour depicts von Mises stress (S. Mises), which concentrates on the skeleton’s edges. (c) Blocked force as a function of the negative pressure under different positive pressures . (d-f) Force-contraction relations when equals +30 kPa, +60 kPa, and +90 kPa, respectively.
Fig. 5: Two analytical approaches to the output force. (a) Approach I: cut the actuator along the center line of the void. The pushing force generated by functions as the drive force for contraction. (b) Tension force as a function of contraction ratio. (c) Pushing force as a function of contraction ratio. (d) Approach II: cut the actuator along the center line of the skeleton pouch. The pushing force generated by functions as resistance to contraction. (e) Tension force as a function of contraction ratio. (f) Pushing force as a function of contraction ratio.
Fig. 6: Dynamic properties of the linear IN-FOAM. The lightweight actuator (size: , in the initial state; weight: 7.46 g; skeleton length: 80 mm) is made of TPU-coated nylon fabrics with a thickness of 0.09 mm. The applied pressures were and . (a-c) Actuation process in the power density tests under positive pressure, negative pressure, and simultaneous positive-negative pressure, respectively. (d-f) Dynamic response of the linear IN-FOAM. Pos.: positive-pressure actuation; Neg.: negative-pressure actuation; Simu.: simultaneous positive and negative pressure actuation. (d) Displacement variations in the power density tests. Under the simultaneous actuation, the actuator can lift a 1 kg load for 3.3 cm within 0.2 s. (e) Power density curves during the lifting process. A peak power density of 0.499 kW kg -1 and an average power density of 0.272 kW kg -1 was achieved in the simultaneous actuation. (f) Mag.: magnitude. Frequency response from target actuation pressure (unit: kPa) to contraction of the actuator (unit: mm). Sinusoidal chirp pressures were generated to drive the actuator. The bandwidths of the linear IN-FOAM under the positive and negative pressure actuation were 0.38 Hz and 0.78 Hz, respectively. (g) Repetitive displacement in the 1st, 1000th, 2000th, 3000th, 4000th, and 5000th cycles (a 1-kg load was applied). (h) Average displacement curve in the first 50 cycles.
Fig. 7: Linear IN-FOAMs with multilayer skeletons. (a) A stacked skeleton with multiple layers can increase the maximum contraction ratio. Assuming the initial length of skeleton pouches is fixed, as the inflated height of skeleton columns increases, the length of the skin that can be compressed into the voids becomes larger. Therefore, the maximum contraction ratio increases. (b) Fabrication of linear IN-FOAMs with multilayer skeletons. The preliminary steps are patterning and then heat-pressing the sheet materials, which are shown in Fig. 1 (e). (c) A four-layer skeleton with a parallel-pouch pattern has inflated beams on both sides to keep the skeleton columns perpendicular to the contraction direction. With the resistance of the inflated beams, the maximum contraction ratio of the actuator (size: ; weight: 48.1 g; skeleton length: 15 cm) is 26 %, consisting of 13 % and 13 % under the positive and negative pressure actuation, respectively. (d) A four-layer skeleton with a zigzag pattern enables the actuator (size: ; weight: 44.3 g; skeleton length: 18.3 cm) to achieve a maximum contraction ratio of 54 %, consisting of 29 % and 25 % under the positive and negative pressure actuation, respectively.
Fig. 8: Skeleton collapse during the contraction of the IN-FOAM with a multilayer parallel-pouch skeleton. (a) Schematic of skeleton inflation (b) Schematic of skeleton collapse under asymmetric tension in the skin during negative pressure actuation. In the experiment, the skeleton may collapse in different directions, not aligned in order, as this illustration shows. (c) Skeleton inflation of an IN-FOAM with a multilayer parallel-pouch skeleton. (d) Skeleton collapse of the multi-layer IN-FOAM.
Fig. 9: A multi-mode IN-FOAM with five motion modes. Motion modes were tested with a 100 g load hung below the IN-FOAM (size: ; skeleton length: 18 cm; weight: 31.5 g). The actuation pressures were set to and . (a) Pneumatic circuit of the multi-mode IN-FOAM. There are six independent channels in the two-layer skeleton, which are denoted by Channels 1-6. The void between the skin and the skeleton is denoted by Channel 7. (b) The actuator bends forward with Channels 2, 4, 5, and 6 inflated. Then it bends further after applying vacuum to Channel 7. (c) The actuator bends leftward with Channels 2, 3, 5, and 6 inflated. After applying the vacuum, it bends further. (d) The actuator performs linear contraction with Channels 2 and 5 inflated, and performs a further contraction under the vacuum. (e) The actuator bends backwards with Channels 1, 2, 3, and 5 inflated, and then bends further under the vacuum. (f) The actuator bends rightwards with Channels 1, 2, 4, and 5 inflated, and then bends further under the vacuum.
| 36.3 % | 51.7 % | 55.3 % | 56.9 % | 57.8 % | |
| 36.3 % | 61.1 % | 65.6 % | 67.4 % | 68.5 % | |
| 36.3 % | 67.4 % | 72.0 % | 73.8 % | 74.8 % | |
| 36.3 % | 72.0 % | 76.4 % | 78.1 % | 79.0 % | |
| 36.3 % | 75.4 % | 79.6 % | 81.2 % | 82.1 % | |
| 36.3 % |
TABLE I: Theoretical values of maximum contraction ratio for the linear IN-FOAMs with multilayer skeletons.
Fig. 10: A crawling robot actuated by a multi-mode IN-FOAM. The actuator’s pneumatic circuits and multi-mode motions are shown in Fig. 9 . (a) Structure of the robot (size: ; weight: 144 g). Both ends of the actuator are clamped in 3D-printed fixtures. Barbed feet are stuck under the fixtures to achieve unidirectional resistance during the robot’s motions. When unactuated, the IN-FOAM is flat, and the robot can be folded up for portability. In the working process, the robot can be unfolded by skeleton inflation. (b-d) The robot moves forward, turns left, and turns right on flat surfaces. Corresponding pressure sequences in different channels follow snapshots of the robot’s motions.
Fig. 11: A soft haptic sleeve actuated by a multi-mode IN-FOAM. (a) Structure of the sleeve (main-body size: ; weight: 51.5 g). The sleeve is designed to be fastened around the wrist via hook-and-loop straps. (b) Pneumatic circuit of the multi-mode IN-FOAM employed in the sleeve. (c-f) The sleeve can convey multiple kinesthetic cues to the wrist by inflating different channels: wrist extension (c); wrist radial deviation (d); wrist extension with radial deviation (e); wrist flexion (f).
| Reference | Low-profile (initial state) | Flexibility | Motion programmability | Maximum contraction ratio (%) | Maximum actuation stress (MPa) | Maximum force-to-weight ratio ( ) | Specific force-to-weight ratio ( ) | Operation pressure |
| IN-FOAMs (this work) | Yes | High | 1D/2D contraction Bending Twisting Rotation | 54 % | 0.108 (+100 kPa -10 kPa) | 9.59 | 0.107 (+90 kPa -0 kPa) | 0 +100 kPa 0 -100 kPa |
| FOAMs [ 23 ] | No | Low | 1D/2D/3D contraction Bending Twisting Rotation | 90 % | 0.64 (-90 kPa) | 10 | 0.1* (-100 kPa) | 0 -100 kPa |
| Pouch Motors [ 17 ] | Yes | High | 1D contraction Bending Rotation | 28 % | 0.092* (+40 kPa) | Unknown | Unknown | 0 +40 kPa |
| Pleated PAMs [ 34 ] | No | High | 1D contraction | 38.2 % | 0.618 (+400 kPa) | 50.2 | 0.126* (+400 kPa) | +2 +400 kPa |
| McKibben Muscles [ 15 , 35 ] | No | High | 1D contraction | 30 % | 1.3 (+500 kPa) | 13* | 0.043* (+300 kPa) | +100 +500 kPa |
| APAMs [ 28 ] | No | Low | 1D contraction Bending | 93.1 %* | 0.035* (-20 kPa +13.79 kPa) | Unknown | Unknown | -20 +10.3 kPa 0 +20 kPa |
TABLE AI: Comparison of pneumatic artificial muscles.
Appendix figures & tables8 assets
Supplementary material from the paper’s appendix.
Appendix
Fig. A1: Compressive resistance measurement of the linear IN-FOAM. (a) Schematic of the compressive resistance measurement method. (b) Experiment setup for the compressive resistance measurements. (c) Relations between the compressive resistance of an inflated IN-FOAM and the displacement.
| Parameter | Value |
| Skeleton size | |
| Skin size | |
| Young’s modulus | 400 MPa |
| Poisson’s ratio | 0.3 |
| Initial plastic strain | 0.04 |
| Contact stiffness | 1 |
TABLE AII: Parameters in finite element (FE) simulations.
Fig. A2: Finite element simulations of IN-FOAMs. (a-d) Cross-sectional views of the simulation results corresponding to Models A, B, C, and D, respectively. Shallow colors are used for the clarity of the cross-sections. (e) Cross-sectional views when the negative pressure increases, with the positive pressure fixed at +30 kPa. (f) Cross-sectional views when the positive pressure increases, with the negative pressure fixed at -60 kPa. (g-i) Simulation results of the IN-FOAM with a flasher-origami-like skeleton. The overall motion and wrinkles in our simulation resemble the experimental results (Fig. 2 (d)).
Fig. A3: Pressure control system built in this work. (a) Schematic of the pressure control system. The Arduino board sends PWM control signals (green dashed lines) to the valves. The pressures in pneumatic circuits are measured by sensors, and the signals are transmitted to the Arduino board (orange dashed lines) as feedback. (b) Control block diagram of the system. Definitions of the symbols: -target pressure, -error between target and actual pressure, -flow rate, -actual pressure, -output voltage of the sensors. (c) A photo of the pressure control system. Labels: #1-Arduino microcontroller board, #2-relays, #3-power module, #4-pressure sensors, #5-diaphragm pump, #6-negative pressure manifold, #7-positive pressure manifold, #8-solenoid valves for pressure-supply control, #9-push-to-connect tube fittings to pressure sensors, #10-push-to-connect tube fittings to external loads, #11-solenoid valves for vent control. (d) Measured negative pressure curves during the force-contraction tests. The positive pressure was kept around +60 kPa, and the negative pressures fluctuated within ±1 kPa of the target values under closed-loop control. (e) Measured positive pressure curves during the force-contraction tests. The negative pressure was kept around -40 kPa, and the positive pressures fluctuated within ±1 kPa of the target values under closed-loop control.
| Reference | Haptic stimuli | Number of motion modes | Low-profile | Number of actuators | Weight of device | Actuator material | Sleeve material |
| This work | I/K | 4 | Yes | 4 | 52 g | Heat-sealable fabric | Heat-sealable farbic |
| [ 36 ] | C/I | No | 8 | 440 g | Elastomeric resin | Knit fabric | |
| [ 37 ] | C/I | Yes | 40 | 67 g | TPU film | TPU film | |
| [ 38 ] | I | Yes | 6 | 40 g | Heat-sealable fabric | Heat-sealable fabric | |
| [ 39 ] | I/T | No | 5/3 | Unknown | Unknown | Fabric | |
| [ 40 ] | C/S/V | No | 6 | 26 g | Woven fabric; latex tube | Knit fabric |
TABLE AIII: Comparison of pneumatic sleeve haptic wearables.
Fig. A4: energy conversion efficiency tests of the linear IN-FOAM. Pos.: positive pressure actuation; Neg.: negative pressure actuation; Neg. clos.: negative pressure actuation with closed-skeleton; Trans.: transferring air from the skin to the skeleton. (a) Schematic of the experiment setup in energy conversion efficiency tests. (b) Relations between input and output work under different operation conditions (the initial amount of air in the closed chambers is 150 mL). (c) Relations between input and output work as the amount of air in the closed-skeleton changes. (d) Relations between input and output work when the actuator is actuated by positive pressure, negative pressure, and transferring air from the skin to the skeleton. (e) Average mechanical energy conversion efficiencies during the actuator’s contraction under different operation conditions (the initial amount of air in the closed chambers is 150 mL). (f) Average mechanical energy conversion efficiencies during the actuator’s contraction as the amount of air in the closed-skeleton changes. (g) Average mechanical energy conversion efficiencies during the actuator’s contraction when the actuator is actuated by positive pressure, negative pressure, and transferring air from the skin to the skeleton.
Fig. A5: Static property of the linear IN-FOAM with closed skeleton. (a) Illustration of the skin pressing to the closed skeleton under negative pressure . : the initial value of before closing the skeleton and applying vacuum to the skin. (b) Blocked force as a function of under different . (c) Variation of the positive pressure in the closed skeleton as increases. (d-f) Force-contraction curves when equals +30 kPa, +60 kPa, and +90 kPa, respectively.
Fig. A6: Multiple operations of IN-FOAMs. By this classification method, there are totally 150 operation modes: . (a) Code definitions of different operation conditions including airflow, connectivity of the skin and the skeleton, and pressurizing sequences. (b) Initial state of the actuator and one example of the operation code. The initial state is that the inner and outer chambers are open to ambient air. The code IP1-OV2-N means that the inner chamber is inflated first, the outer chamber is secondly driven by vacuum, and the inner and outer chambers are not connected. (c) Operation conditions taking connectivity and pressurizing sequences into consideration. (d) Operation conditions when the skin and the skeleton chambers are independent and driven simultaneously.