Parallel Fin Ray Soft Gripper with Embedded Mechano-Optical Force Sensor

Abstract

The rapid advancement in soft robotics over the past decade has driven innovation across the industrial, medical, and agricultural sectors. Among various soft robotic designs, Fin Ray-inspired soft grippers have demonstrated remarkable adaptability and efficiency in handling delicate objects. However, the integration of force sensors in soft grippers remains a significant challenge, as conventional rigid sensors compromise the inherent flexibility and compliance of soft robotic systems. This study presents a parallel soft gripper based on the Fin Ray effect, incorporating an embedded mechano-optical force sensor capable of providing linear force measurements up to 150 N. The gripper is entirely 3D printed using thermoplastic elastomers (TPEs), ensuring a cost-effective, scalable, and versatile design. The proposed sensor architecture leverages a gyroid lattice structure, yielding a near-linear response with an $R^2$ value of 0.96 across two force regions. This study contributes to the development of sensorized soft grippers with improved force-sensing capabilities while preserving the advantages of soft robotic manipulators.

1. Introduction

The advancement in soft robotics in various sectors, such as the industrial [1,2,3,4,5], medical [6,7], and agricultural [8] domains, is consolidating at the beginning of this century. This type of robotics, which uses soft materials like silicone, rubber, or thermoplastic elastomers (TPEs), is characterized by its ability to adapt to various objects, interact with them without causing damage, and enable safe interactions in robot–environment, robot–robot and robot–human scenarios.

One of the standout applications of this type of robotics is object manipulation, particularly through the use of soft grippers. In the literature, various designs of these robotic grippers have been proposed [9,10,11,12], employing different geometries, materials, and actuation methods to achieve adaptable, firm, and non-damaging robot–object interactions. The design of these grippers often draws inspiration from nature, such as those that mimic the grip of octopuses [13,14] or geckos [15,16].

However, a major challenge in soft robotics remains the integration of force sensors capable of real-time force measurement without compromising the inherent flexibility and compliance of soft materials. Sensorization in soft robotics is still an emerging field, and most studies have relied on sensors originally designed for traditional robotics. These sensors, being rigid, are often incompatible with the flexibility of soft materials [17,18]. This mismatch undermines the core advantages of soft robotics, such as their ability to distribute forces uniformly and conform to objects’ shapes for improved adaptability. To overcome this limitation, researchers have explored alternative sensing methods compatible with the elastic and flexible properties of soft materials, leveraging technologies such as magnetics [19], acoustics [20], and textile-based sensors [21].

One of the most widely recognized designs in soft robotics is based on the Fin Ray effect [22,23,24,25,26], inspired by the anatomy of fish fins, which exhibit both flexibility and adaptability. Fin Ray-inspired grippers, however, often lack integrated force sensors due to technical difficulties [22,23,24,25], as shown in Figure 1. Although some studies have attempted to incorporate sensors into these grippers [26], they frequently rely on conventional rigid sensors that compromise the gripper’s ability to maintain its inherent adaptability. This adaptability is crucial for distributing forces over larger contact areas and conforming to objects without requiring active position control [27]. Therefore, advancing sensor technologies that seamlessly integrate with the soft and flexible nature of these materials remains a key focus in soft robotics research.

For this reason, this article proposes a Fin Ray-inspired parallel soft gripper with an integrated mechano-optical force sensor, designed to provide real-time force measurement while preserving adaptability. Unlike conventional soft grippers that rely on external force sensors or resistive/capacitive sensing elements, our approach achieves direct force sensing integration within the gripper structure itself. The gripper is entirely 3D printed using thermoplastic elastomer (TPE) filaments, ensuring a cost-effective, scalable, and versatile design. By leveraging a gyroid lattice structure, the integrated sensor achieves a near-linear force response, enabling precise contact force control in robotic manipulation. Compared to existing soft grippers, which often suffer from nonlinear sensing behavior, external sensor placement limitations, and high manufacturing costs, our design provides an all-in-one solution that combines structural adaptability with embedded force-sensing capability. The proposed soft gripper, with its adaptable gyroid lattice structure and integrated force sensor, allows for the gentle grasping of fragile objects, such as small fruits or vegetables, minimizing the risk of bruising. Consequently, the gripper is particularly suited to applications that require both adaptability and delicate force regulation, such as small fruit harvesting in agriculture or precision pick-and-place operations in industry.

The remainder of the article is organized as follows. Section 2 provides a brief state of the art on sensors embedded in soft grippers, while Section 3 details the steps followed for the design and manufacturing of the proposed soft gripper with embedded mechano-optical force sensor. A characterization and experimental evaluation of the designed soft robotic fingers and the integrated force sensor are presented in Section 4. Finally, Section 5 summarizes major conclusions.

2. State of the Art

Like other soft robotic systems, soft grippers are characterized by their intrinsic deformability and compliance [28], providing significant advantages in adaptability and safety when interacting with dynamic and unpredictable environments. However, they also present challenges in terms of sensorization and control due to their high inherent number of Degrees of Freedom (DoFs). The complexity of soft robotic finger control is compounded by the need for precise feedback on parameters such as force, deformation, and position during operation.

To address these challenges, sensor integration has emerged as a pivotal area of research in soft robotics. Integrating soft sensors into actuators enables real-time monitoring and adaptation to changing conditions, thereby enhancing both control and functionality. Unlike traditional rigid sensors, soft sensors are specifically designed to provide flexibility, adaptability, and the ability to conform seamlessly to soft robotic systems. These sensors measure physical parameters such as pressure, deformation, and temperature while actively contributing to the real-time control and performance of soft robotic fingers. Recent advancements, including liquid metal-filled microchannels and 3D-printed resistive sensors, have demonstrated significant potential for improving the versatility and precision of soft grippers, particularly in applications involving delicate or irregular objects [29,30].

One of the significant advancements in soft robotics is the development of 3D-printed resistive soft sensors that can be co-fabricated with the robot bodies themselves, as demonstrated by [29]. This method enables the creation of integrated, multifunctional soft robotic actuators with enhanced sensing capabilities by incorporating sensors directly into the actuator’s structure, providing greater design flexibility and customization. The co-fabrication technique also allows for the production of lightweight and flexible sensors, essential for maintaining the compliance of soft robotic systems.

Additionally, soft sensors utilizing liquid metal-filled microchannels offer unique advantages in terms of compliance and durability. These sensors can stretch and bend with the soft robotic finger’s movements without compromising functionality, making them ideal for dynamic applications [30]. Similarly, embedded barometric sensors can measure internal pressure changes, allowing for the precise control of soft robotic fingers by monitoring deformation and exerted forces. These sensors are cost-effective and provide reliable data, crucial for precise control and feedback in soft robotics. A review in [31], shown in Figure 2, illustrates the range of these soft sensor technologies, highlighting their potential and versatility in the field.

Therefore, various sensing technologies have been explored in soft robotics to provide tactile and force feedback. Piezoresistive sensors are valued for their low cost and ease of integration, but they suffer from hysteresis and low spatial resolution [43]. Piezoelectric sensors offer a high dynamic range but are unable to measure static forces [44]. Capacitive sensors provide high spatial resolution but require complex filtering electronics and are prone to noise interference [45]. On the other hand, optical fiber-based sensors are immune to electromagnetic interference and are chemically inert, but their integration into soft robotic devices remains limited due to the requirement for external signal processing systems [46]. More advanced approaches, such as Gelsight and Contactile, enable detailed deformation and slip detection; however, their reliance on rigid structural components makes their implementation in flexible grippers challenging [47,48].

Recent research has driven the development of 3D-printed mechano-optical force sensors for soft robotic grippers, offering a precise, cost-effective alternative for force sensing [49]. These sensors operate by modulating optical intensity in response to mechanical deformation, with changes measured by a photodetector. They demonstrate a wide dynamic force range (0–16 N), fine resolution (0.015 N), linear response, and excellent durability (>100,000 cycles), all achieved in a single-step 3D-printing process. This integration simplifies manufacturing while allowing customized force sensing characteristics through adjustments in material properties and lattice geometry.

Due to their adaptability, modularity, robustness, and cost-effectiveness, these sensors are particularly well suited to applications demanding real-time force measurement and embedded integration into soft robotic systems. In agriculture, they facilitate the gentle harvesting of delicate fruits and vegetables while preventing mechanical damage. In industrial automation, they enable precise force control in pick-and-place operations to minimize handling-induced damage. Additionally, their miniaturization potential and direct soft structure integration make them viable for haptic interfaces and medical applications, including surgical tools and wearable force monitoring systems for rehabilitation and assistive robotics. These capabilities position the proposed sensor as a versatile and effective alternative within the expanding field of soft robotics sensing.

3. Parallel Fin Ray Soft Gripper Design

Robotic grippers can be designed with various geometries and actuation methods. Among these, parallel grippers are the most common due to their ease of control, suitability for repetitive tasks, cost-effectiveness, and reliability. As a result, they dominate industrial process automation. In this regard, within soft robotics, there are various designs of adaptable parallel-type soft grippers. This stands out compared to another significant group within soft grippers, pneumatically actuated ones, which are more sensitive to potential manufacturing defects and dependent on a constant pneumatic supply line. This reduces their efficiency, as it requires converting electrical energy into the potential energy of compressed air, which increases costs compared to grippers that only need electrical energy.

Among adaptable soft grippers, those that leverage the Fin Ray effect stand out for their reliability, ease of manufacturing, and simple control. This bioinspired effect, discovered by biologist Leif Kniese, mimics the deformation behavior of fish fins. The structure of the actuators that make up these grippers consists of two alternating tension and compression sides flexibly connected by rigid ribs. This structure enables a high degree of movement with minimal force input [50].

Building upon these principles, we designed a parallel soft gripper that not only leverages the Fin Ray effect but also integrates a mechano-optical force sensor for embedded force measurement. To achieve this, we selected a compact and monolithic actuator design (see Figure 3), ensuring scalability and modularity, both critical for industrial applications. To optimize the structural parameters of the Fin Ray fingers, we followed the methodology outlined in [24], which highlights the importance of rib configuration in tuning both stiffness and displacement. A maximum applied force of 150 N was established as the design constraint to ensure that the deformation remained within an optimal range. The ribs were iteratively adjusted to balance mechanical compliance and structural integrity, allowing the gripper to maintain adaptability while preventing excessive deformation under high loads.

The gripper was fabricated using two different materials: thermoplastic elastomer (TPE) with a hardness of 70A and polylactic acid (PLA). TPE was selected for the robotic fingers and embedded sensors, as these components directly interact with objects, requiring both flexibility and durability. In contrast, PLA was chosen for structural components that connect the robotic fingers to the wrist of the robotic manipulator, ensuring mechanical stability without interfering with the gripping mechanism. The use of these materials aligns with Fused Filament Fabrication (FFF) technology, enabling the production of complex geometries while maintaining cost-effectiveness.

The design of the embedded mechano-optical force sensor was guided by the study conducted in [49], which analyzed the mechanical response of various metamaterial lattice structures, including simple cubic, face-centered cubic (FCC), octahedral, body-centered cubic (BCC), 2.5D X-cell, Schwartz D, and gyroid lattices. Among these, both gyroid and BCC lattices were identified as isotropic, meaning they provide uniform resistance and deformation along all three axes. This isotropic nature is crucial for force sensing applications, as it ensures predictable and repeatable deformation behavior, which is fundamental for achieving a linear force response in soft robotic applications.

The gyroid lattice was ultimately selected due to its favorable balance between stiffness and deformation characteristics. Specifically, it exhibits a gradual transition from a low-stiffness regime to a high-stiffness regime under moderate deformation levels. This property allows for high sensitivity in the initial compression range while maintaining structural integrity under higher forces. Additionally, its continuous, triply periodic minimal surface (TPMS) geometry ensures even stress distribution, unlike traditional strut-based structures such as BCC or FCC, which often concentrate stresses at node intersections. This stress distribution enhances both durability and repeatability.

Moreover, the gyroid lattice offers greater tunability in stiffness and deformation response by adjusting unit cell size and infill density, as demonstrated in the experimental results. This tunability is critical for customizing sensor performance across various force measurement ranges in robotic applications. Furthermore, its superior manufacturability in flexible filament 3D printing ensures reliable fabrication while maintaining mechanical integrity, making it the optimal choice for integration into soft robotic grippers.

As for the sensor structure, shown in Figure 4, it comprises three main parts: the primary body, which houses both the infrared (IR) photodiode and the IR emitting LED, connected by a channel; the compression layer, containing the gyroid pattern, which provides the sensor with the ability to linearly dampen the load; and the contact layer, which features an internal protrusion that occludes the channel in the main body connecting the diodes, cushioned by the compression layer. This change in the degree of occlusion of the infrared light emitted by the LED is read by the IR photodiode and processed by a microcontroller, which converts the analog signal into a calibrated force value.

Manufacturing

For the manufacturing of the soft robotic fingers that constitute the parallel soft gripper, a 3D printer, specifically a modified Ender 3, was used. This modification, in the form of a direct drive system, was adapted to process flexible TPE filament, which is challenging due to its hyperelasticity. The printing temperature was set to 225 °C, allowing proper extrusion of the filament while preventing clogging in the printer’s hotend or under-extrusion. Additionally, the extrusion multiplier was iteratively adjusted to achieve a continuous filament flow, ensuring a uniform and consistent print quality for both the soft robotic finger and the embedded sensor.

To evaluate the reliability and repeatability of the manufacturing process, key parameters related to the printer’s accuracy and precision were considered. The Creality Ender 3 features a positioning accuracy of ±0.1 mm on the X/Y axes and ±0.05 mm on the Z axis, with a layer resolution adjustable between 0.05 and 0.4 mm. The dimensional accuracy for printed parts is approximately ±0.2 mm. Regarding repeatability, the system maintains a variation of ±0.1 mm on the X/Y axes and ±0.05 mm on the Z axis, influenced by belt tension and mechanical stability. These tolerances ensure consistency in multiple prints, reducing variations in sensor performance. To achieve proper layer adhesion and reproducibility, the parameters presented in

Table 1. Printing parameters for flexible TPE 70A and PLA filaments.

Parameter	TPE Value	PLA Value	Unit
Nozzle diameter	0.4	0.4	mm
Layer height	0.2	0.2	mm
Infill	100	20	%
Material temp	225	200	°C
Build plate temp	50	50	°C
Print speed	20	50	mm/s
Extrusion multiplier	115	100	%

were utilized.

When creating the gyroid-type pattern, an infill of 15% was selected (see Figure 5). This percentage was determined after several iterations aimed at achieving a sensor capable of operating in a low-force range, allowing for the manipulation of fragile objects, such as small fruits or vegetables, without causing bruising. This infill pattern is available in most slicer software, such as Ultimaker Cura. For its proper manufacturing, it is necessary to remove the outer layers of the structure that will serve as the interface.

To implement the proposed parallel soft gripper on a commercial cobot, adapters were designed to attach the soft robotic fingers to the YuMi end effector, shown in Figure 3c. These adapters replace the YuMi’s rigid fingers and allow for quick changes between fingers when needed. The adapters were manufactured in rigid plastic, specifically PLA filament.

4. Characterization and Experimental Results

The characterization of the mechano-optical force sensor (Figure 6) was conducted using the same dimensions and gyroid infill pattern (15%) as in the final gripper design. The experimental setup included a dial indicator to measure the deformation distance and an Imada SSII-20R dynamometer to apply controlled compression to the sensor’s contact surface (Figure 6a). After calibrating the force sensor with the dynamometer, multiple load cycles were performed to assess repeatability and hysteresis effects. The results confirmed that deviations remained within acceptable limits, ensuring the reliability of the sensor readings. The sensor exhibited a hysteresis of 11.92%, determined through multiple loading and unloading cycles at a rate of 15 mm min⁻¹. This value is consistent with the results reported in [49]. In addition, the maximum sample standard deviation was found to be 0.5 for the voltage-deformation measurement (Figure 6c) and 3.3 for the force-deformation measurement (Figure 6d). The sensor demonstrated a sensitivity of 0.004 V 0.16 N, representing the smallest detectable change in force.

The force-response behavior of the gyroid-patterned sensor closely aligns with the findings of [49] (Figure 6c,d). The response is divided into two distinct, nearly linear regions, both exhibiting a high coefficient of determination ($R^2$ = 0.96) (Figure 6e,f). In the first region (0–2.5 mm), the gyroid lattice maintains its structural integrity and exhibits elastic resistance, supporting forces up to approximately 24 N while preserving a linear response. However, beyond 2.5 mm compression, the gyroid structure collapses, leading to a shift in mechanical behavior. In this second region (2.5–4 mm), force measurements are primarily dictated by the compression of the deformed TPE layers, enabling the sensor to reach a peak force of approximately 150 N. As shown in Figure 6b, this transition occurs precisely at 2.5 mm compression due to the complete structural collapse of the gyroid pattern, fundamentally altering the sensor’s mechanical response.

After characterizing the mechano-optical sensor, the parallel soft gripper was mounted onto the HortiRobot dual-arm robotic system. The HortiRobot experimental platform [51] is based on an ABB YuMi®—IRB 14000, a collaborative robot with two high-dexterity seven-axis arms [52].

To evaluate the sensor’s performance, we tested eight different geometries, each measuring 20 × 20 × 20 mm (Figure 7).

The grasping test evaluated sensor response variations when an object was offset from the gripper’s center. Each geometry produced distinct force readings, with notable differences in semi-sphere and pyramid-shaped objects. The semi-sphere’s curved surface caused deeper indentation due to its smaller force application area, whereas the flat surface exhibited lower contact force readings due to a larger contact area.

As for the pyramid, it rotated along its axis during manipulation. This showed that contact on one finger was concentrated along the edge of a face, while the other finger made contact with a flat face of the pyramid. As in the previous case, the sensor readings varied depending on the contact area. In this specific scenario, where the manipulated object was stiffer than the gripper, the differences in sensor responses were particularly notable. Tests have also been conducted to measure the maximum load lifted. Since the robotic system can only lift up to 250 g, including the gripper, the tests have been limited to this maximum load, which has been successfully lifted.

Therefore, these tests highlight not only the sensitivity of the embedded sensors within the soft gripper but also their adaptability to different object geometries. The gripper’s compliance enables a secure grasp while minimizing the risk of damage, as its structure dynamically reshapes to match the contours of the object. This property allows the system to detect and measure forces even when they are localized or asymmetrically distributed. When such forces are exerted, the deformation of the soft fingers redistributes the contact area, enabling the sensor to capture the resulting force variations effectively. This observation underscores the importance of incorporating foldable patterns, particularly the gyroid-type lattice, in soft sensors manufactured with materials such as TPE 70A. The gyroid structure facilitates uniform force distribution and exhibits a quasi-linear response, making it highly suitable for integration into soft robotic grippers.

These results further demonstrate the effectiveness of the proposed mechano-optical sensor in providing accurate force feedback while preserving the compliance of the soft gripper. Compared to capacitive or resistive soft sensors, this approach offers a cost-effective and scalable solution. Moreover, the gyroid-patterned sensor structure minimizes deformation artifacts, which are commonly observed in traditional silicone-based force sensors, making it a promising candidate for robotic applications requiring embedded sensing with minimal material constraints.

Table 2. Key performance metrics of the proposed mechano-optic sensor in comparison with previously reported force and pressure sensors.

Sensor	Type of Sensing	Fabrication Steps	Sensitivity	Linearity (${\mathit{R}}^2$)	Durability (Cycles)	Operation Range
Mechano-optic sensor	Mechanical deformation- induced optical intensity modulation	1	0.16 N	$0.96-0.96$	10 000	0–150 N
Optoelectronic [32]	Stretchable optical waveguide	>8	0.4 N dB	-	-	20 N
Polymer hydrogel sensor [53]	Variation in resistance of the hydrogel under strain	>10	Deformation of 12 $\mathsf{\mu }$m	$0.98$	2000	300% strain
Somato-sensor [54]	Resistive sensor	>4	-	-	-	80 kPa
Shielded liquid metal silicone sensor [55]	Capacitive	>15	0.003 N	-	1000	>20 N
Multi-direction flex sensor [56]	Resistive	>5	-	-	<100	5 N
3D printed haptic sensor [57]	Resistive	>3	-	$0.9728$	-	0–13 N
Triaxial force sensor [58]	Capacitive	>8	0.0074 N	0.93–0.98	-	0–0.25 N
Triboelectric nanogenerator pressure sensor [59]	Triboelectric nanogenerator	>8	12.61 pF· kPa−1	$0.99213$	-	5 kPa
Ultrastretchable hydrogel sensor [60]	Resistive sensor using liquid metal	>10	0.25 kPa	-	1000	150 kPa
Ionic liquid-based force sensor [61]	Resistance change in ionic liquid	>10	-	-	-	12 N

presents a comparative analysis of key characteristics between the designed mechano-optical sensor and various previously reported force and pressure sensors. In addition to exhibiting high sensitivity (0.16 N) and strong linearity ($R^2$ = 0.96), our sensor provides several notable advantages: (1) a simplified fabrication process with reduced manufacturing costs; (2) high durability, withstanding over 10,000 cycles without significant degradation; (3) an extensive force measurement range (0–150 N), enabling adaptability to different grasping tasks. These attributes contribute to the seamless integration of this sensor into the proposed soft gripper, allowing it to measure contact forces while maintaining the structural adaptability and surface reshaping capabilities characteristic of soft robotics.

5. Conclusions

The development of soft robotics has significantly advanced across various sectors, including the industrial, medical, and agricultural domains. However, integrating force sensors into soft grippers remains challenging, as conventional sensors often compromise the flexibility and compliance of soft robotic systems. This study presents a parallel Fin Ray soft gripper with an embedded mechano-optical force sensor, enabling real-time force measurement without compromising adaptability. This proposed design aims to address the critical need for real-time sensing in soft grippers without compromising their flexibility, adaptability, or functionality.

The proposed gripper, entirely fabricated using 3D-printing technology with PLA and TPE materials, is cost-effective, scalable, and adaptable, making it well suited to diverse robotic applications. Additive manufacturing enables precise customization and rapid prototyping, making the design highly adaptable to a wide range of applications. A key innovation is the integration of a mechano-optical force sensor directly within the gripper’s Fin Ray structure, providing a compact and functional sensing mechanism for enhanced control and feedback during operation. The use of a gyroid lattice pattern was specifically studied, yielding a linear force-response behavior with an $R^2$ value of 0.96, capable of accurately measuring forces ranging from 0 N to 150 N. This combination of features makes the gripper particularly suited to tasks requiring precise contact force control, such as small fruit harvesting in the agricultural sector or pick-and-place operations in industry.

Compared to other soft grippers with integrated force sensors, the proposed design offers a fully 3D-printed structure and demonstrates higher sensitivity in force measurements while preserving adaptability. Capacitive and resistive sensors are widely reported due to their widespread adoption [62,63,64,65], but they face challenges related to hysteresis, limited dynamic range, and complex fabrication process. Many of these soft sensors require embedded conductive elements or multilayer manufacturing, whereas the mechano-optical sensor leverages a simpler fabrication process and provides a scalable solution. Although external wiring is necessary for power and data transmission, the optical sensing method minimizes additional electronics inside the soft structure, maintaining flexibility and compliance. By leveraging a gyroid-patterned lattice for force distribution, this approach provides a reliable and efficient sensing mechanism for soft robotic applications.

Future work will focus on accurately modeling soft gripper joints in the Robot Operating System (ROS) to improve simulation and real-world performance. Additionally, we will explore trajectory planning for the HortiRobot dual-arm platform, optimizing its integration for agricultural harvesting applications.