Introduction: Fiber Bragg grating (FBG) optical sensors can capture precise physical phenomena based on how their Bragg wavelengths shift due to interactions with temperature and strain. The small size and highly bendable capability of an FBG sensor make it an attractive sensing solution in a wide range of applications. Despite these advantages, FBG optical sensors are not stretchable and therefore are not frequently used in soft robotics. Recently, research has shown that FBG sensors can unfold with a stretchable material when embedded in a sinusoidal manner, this allows for their use in soft robotics (Li. Xu et al., "Stretchable fiber-Bragg-grating-based sensor"). FBG sensors can then inform the soft robotic system of its environment while maintaining the flexibility and softness of the soft robot. In this work, we have designed a collapsible soft robotic muscle and explored how changing strain due to different loads sensed by an FBG sensor can be used to inform a soft robotic muscle of an object's weight and actuate the soft robotic muscle correspondently.
Materials and
Methods: The soft robotic muscle is inspired by collapsible cups and is designed to collapse into a disk shape when vacuumed. This is significant as the muscle collapses internally, unlike similar collapsible soft robots which collapse on themselves and limit shrinkage. (Rogatinsky, J. et al., “A Collapsible Soft Actuator Facilitates Performance in Constrained Environments”). Loads can be placed on the silicone “hammock,” seen in Figure 1, and lifted when vacuumed and lowered when pumped. The collapsible movement is geometry-dependent, caused by circular rings connected to thin walls. The rings are reinforced with Polylactic acid (PLA) rings, ensuring the silicone walls do not cave when vacuumed.
Weight detection is achieved through an FBG embedded 0.2mm into the hammock, measuring 5 x 145 x 1 mm. The hammock, like the rest of the soft robot, is made of 50% Ecoflex-00-30 and 50% DragonSkin-00-30 (see Figure 1). As heavier weights result in larger wavelength shifts, the relationship can be modelled with equations.
Wavelength shift amount is used in the load classification equation in MATLAB, which then determines the intensity of the vacuum necessary for the load to be lifted. Weight classification is organized by ranges rather than exact weights, to achieve an effective control system with minimized error. With vacuuming/pumping intensity determined, the information is sent to an Arduino microcontroller which controls actuation. The wiring schematic is shown in Figure 2, where the Arduino board indicates that both the vacuum and air pump are pulse width modulated (PWM) outputs, allowing for variable intensity activation.
Results, Conclusions, and Discussions: The FBG optical spectrum is measured by a broadband light source launched into the FBG via an optical circulator. Reflected FBG Bragg wavelengths are received by an optical spectrum analyzer through the circulator. Objects of differing weights were placed on the hammock to determine a mathematical relationship between weight and Bragg wavelength shift. Similarly shaped objects were used, minimizing FBG wavelength shifting caused by geometry instead of weight.
The relationship between weights (w) in grams placed on the hammock and wavelength (λ) in nanometers can be represented by equation (1). Subsequent wavelength shifting will decrease optical power at the rising edge (shorter wavelength side), represented by equation (2).
These equations quantify the relationships between the weight and the FBG wavelengths; where increasing weight (w) causes increasing wavelength (λ) and decreasing rising edge (shorter wavelength side) optical power (P) (see Figure 4). The wavelength shift and power change equations have a respective R-square of 0.9352 and 0.9593. Both equations and R-square values were calculated in MATLAB, using 5 different trials of 34 different weights between 0-405g.
Three weights, 10g, 50g, and 100g, were tested for classification using the previous equations and successful actuation of the soft robot. The soft robot lifted and brought all test weights to their highest point through collapsing motion. Figure 5 shows the actuation sequence of the soft robotic muscle, which shrinks by 78%. With weights between 0-405g, a maximum wavelength shift of 0.41 nm is observed. Exact wavelength shift identification is challenging due to the small maximum FBG wavelength shift, lack of precision determining shift amount using an optical spectrum analyzer, and ±0.1 nm FBG wavelength error. This occasionally caused incorrect weight classification. Errors can be overcome by using the more significant change in optical power, seen in Figure 3.
FBG Wavelength readings for weight controlling a soft robotic muscle could potentially increase effectiveness of small-scale load-bearing mechanisms. To further improve the system’s precision, simplicity, and cost, a laser and photodetector system can be used to capture precise real-time changes in optical power, instead of deriving the power change from a static optical spectrum.
Acknowledgements (Optional): This material is based upon work supported by a National Science Foundation Research Experiences for Undergraduates (REU) site program under Grant No. 1950581.
