A TPU-Based 3D Printed Robotic Hand: Design and Its Impact on Human–Robot Interaction

Abstract

This study outlines the design and evaluation of a biomimetic robotic hand tailored for Human–Robot Interaction (HRI), focusing on improvements in tactile fidelity driven by material choice. Thermoplastic polyurethane (TPU) was selected over polylactic acid (PLA) based on its reported elastomeric characteristics and mechanical compliance described in prior literature. Rather than directly matching human skin properties, TPU was perceived as providing a softer and more comfortable tactile interaction compared to rigid PLA. The robotic hand was anatomically reconstructed from an open-source model and integrated with AX-12A and MG90S actuators to simplify wiring and enhance motion precision. A custom PCB, built around an ATmega2560 microcontroller, enables real-time communication with ROS-based upper-level control systems. Angular displacement analysis of repeated gesture motions confirmed the high repeatability and consistency of the system. A repeated-measures user study involving 47 participants was conducted to compare the PLA- and TPU-based prototypes during interactive tasks such as handshakes and gesture commands. The TPU hand received significantly higher ratings in tactile realism, grip satisfaction, and perceived responsiveness (p < 0.05). Qualitative feedback further supported its superior emotional acceptance and comfort. These findings indicate that incorporating TPU in robotic hand design not only enhances mechanical performance but also plays a vital role in promoting emotionally engaging and natural human–robot interactions, making it a promising approach for affective HRI applications.

1. Introduction

As robotic technologies continue to advance in various fields, the discussion of Human–Robot Interaction (HRI) [1,2,3,4] has evolved beyond a sole focus on mechanical precision to increasingly emphasize user experience and emotional acceptance as critical components of interaction design [5]. In scenarios involving physical contact between humans and robots, factors such as appearance, movement, and especially tactile sensation significantly influence user trust and engagement [6,7,8,9]. Among these factors, the robotic hand represents the most intuitive and socially significant interface for interaction. Although traditional robotic hands composed of metal or hard plastic offer high control accuracy, they often evoke a sense of coldness and physical detachment [10]. To address these limitations, advances in soft robotics have introduced materials and structural designs that promote more human-friendly interactions [11,12,13,14]. One particularly promising material is thermoplastic polyurethane (TPU) [15,16,17], which combines flexibility with durability. TPU retains its shape under repeated deformation and offers elastomeric properties that are often perceived as softer and more compliant than rigid plastics during physical interaction.Its compatibility with 3D printing [18,19,20,21] of fused deposition modeling (FDM) has established it as a preferred material to produce robotic skins, soft grippers, and wearable robotic hands [22,23,24,25]. Previous studies have shown that realistic tactile feedback can enhance emotional immersion and trust in HRI settings.

This study presents an upgraded robotic hand based on an existing TPU-based design, which incorporates structural improvements and integrated controller hardware to improve motion consistency and emotional responsiveness [25,26,27]. A comparative experiment is conducted using two identically designed robotic hands—one fabricated from TPU and the other from polylactic acid (PLA)—in order to comparatively assess how differences in material properties influence user perception, emotional response, and preference under identical geometric and control conditions.

The key contributions of this study are as follows:

• Provides an experimental comparison of anthropomorphic robotic hands with identical structural designs but distinct material compositions (TPU vs. PLA), highlighting their differential impact on user experience.
• Employs a hybrid HRI evaluation framework that integrates quantitative measures (e.g., 5-point Likert scales, paired t-tests) with qualitative analyses, including open-ended emotional feedback and suggestions for improvement.
• Demonstrates empirically that user-perceived tactile and emotional characteristics of robotic hand materials significantly influence trust, emotional acceptance, and the overall quality of interaction.

The remainder of this paper is organized as follows. Section 2 reviews related work. Section 3 presents the design and control integration of the enhanced TPU-based robotic hand. Section 4 describes the experimental setup, the user study methodology, and the resulting quantitative and qualitative findings. Section 5 concludes the paper and discusses directions for future research.

2. Related Work

Recent developments in HRI research have expanded beyond functional performance to emphasize user-centered dimensions such as trust, emotional acceptance, and comfort. Goodrich and Schultz [5] identified trust building as a foundational aspect of HRI, stating that robot appearance and behavior play a critical role in fostering emotional engagement. Dragan et al. [6] demonstrated that physical movement cues—specifically motion velocity and the smoothness or fluidity of the trajectory—substantially shape how users perceive both the robot’s safety and its underlying intent. These findings underscore the importance of high-quality physical interaction in shaping the overall HRI outcomes. In this context, tactile perception emerges as a particularly influential element. It encompasses the subjective experience of touch, including sensations of softness and temperature during physical contact. Tactile perception has been increasingly recognized as the key to eliciting emotional responses and forming human-like impressions of robots. However, it has often been regarded as secondary to visual or behavioral factors within the field of HRI.

Traditional robotic hands were initially constructed from rigid materials such as metals and hard plastics to ensure precise control and mechanical durability. However, these designs often conveyed a cold mechanical impression, creating psychological barriers during physical interactions [10]. In response, the field of soft robotics has emerged, driven by innovations in materials and structural design that emphasize human likeness and emotional engagement. Kim et al. [11] introduced soft robotic actuators that emulate biological muscles and skin, highlighting their potential for safer and more intuitive interaction [28]. More recently, TPU has attracted considerable interest as a promising material for HRI applications due to its high elasticity, abrasion resistance, and compliant surface characteristics that are often perceived as comfortable during human–robot contact. Studies such as those by Chen et al. [21] have shown that TPU-based robotic interfaces improve user engagement by offering a more natural and emotionally acceptable tactile experience.

The increasing accessibility of 3D printing technologies, particularly FDM, has enabled researchers to quickly and cost-effectively prototype robotic components using a wide range of materials and structural configurations. This advancement has spurred significant innovation in robotic hand design, especially in the customization of physical properties such as flexibility and surface texture. Choi et al. [25] introduced a TPU-based robotic hand with integrated control electronics, enhancing both movement fluidity and user comfort. Their research emphasized the close relationship between material properties and control integration, demonstrating how these factors jointly affect mechanical performance and user perception. However, their study did not isolate material composition as a variable while keeping structural and mechanical parameters constant. This limitation motivates the present study, which conducts a controlled comparative evaluation using identical geometric designs and control architectures to examine material-related perceptual differences.

Evaluating the effectiveness of HRI requires a multidimensional approach. Dautenhahn [29] proposed a comprehensive evaluation framework that includes both quantitative measures, such as task efficiency and response time, and qualitative dimensions, including emotional comfort, perceived safety, and naturalness of interaction. Although previous studies have applied this framework to behavioral or conversational agents, empirical research specifically focusing on tactile experiences with physically identical robotic hands made from different materials remains limited. This study adopts and adapts the Dautenhahn framework to investigate how material properties influence the user experience. Using a hybrid evaluation method that using Likert scale surveys combined with statistical comparisons and qualitative content analysis of open-ended emotional responses, this research provides empirical insight into material-centric design considerations in HRI by directly linking tactile realism to emotional acceptance and user trust.

This study investigates two foundational concepts in the field of HRI: emotional acceptanceand tactile perception. The following section defines each term in detail and elucidates their relevance to the scope of the present research.

Emotional Acceptance. Emotional acceptance refers to the extent to which users feel emotionally comfortable, at ease, and psychologically safe when interacting with a robot. Dautenhahn [29] identified emotional acceptance as a critical component to foster trust and enhance user immersion in HRI. Fong et al. [30] further emphasized its influence on social acceptance and affective design [31,32,33]. In this study, emotional acceptance is assessed through both qualitative responses and quantitative Likert scale ratings that address questions such as “Does the interaction feel comfortable?”, “Do you feel safe?” and “Do you feel emotionally natural?”

Tactile Perception. Tactile perception refers to the subjective sensory experience of texture, hardness, pressure, and responsiveness arising from physical interaction with a robot. Okamura [34] characterized it as the conscious recognition of mechanical and textural signals, while Culbertson et al. [35] underscored its relevance to user immersion and acceptance in HRI. In this study, tactile perception is evaluated using multidimensional metrics, including satisfaction with surface texture, perceived appropriateness of hardness, naturalness of tactile sensation, and similarity to human touch [34].

3. Implementation

This study empirically examines the impact of 3D printing materials on tactile perception and emotional acceptance in HRI. Two anatomically identical robotic hands were fabricated using different materials: TPU and PLA. By ensuring that material properties were the sole variable, the experimental design enables a clear attribution of any observed differences in user responses to the materials themselves. Participants engaged with each robotic hand through a set of predefined interaction tasks, including handshake, rock-paper-scissors, and a neutral pose. Immediately after each interaction, they completed a structured questionnaire evaluating five key dimensions of HRI quality: motion consistency, grip force satisfaction, response speed, surface texture satisfaction, and hardness satisfaction. In addition, emotional responses and qualitative impressions were collected through open-ended questions to provide a comprehensive understanding of the user experience. This research contributes to the HRI field by offering controlled empirical evidence on how material composition influences users’ perceptions of tactile realism, emotional comfort, and preference. The findings aim to inform the future development of robotic hands designed for social, caregiving and educational applications, where emotional receptivity is essential.

3.1. Robotic Hand Design and Fabrication

This study aims to improve tactile receptivity and emotional engagement in HRI environments by designing a robotic hand using TPU. The primary material selected is the TPU 95A filament, manufactured by Moment Co., Beijing, China with a diameter of 1.75 mm, a dimensional tolerance of ±0.05 mm, and a Shore A hardness of 95. Previous research suggests that Shore A hardness of human skin typically ranges from 20 to 40, serving as an important benchmark for designing biomimetic materials. Although TPU 95A exhibits greater hardness than human skin, it was chosen for its ability to maintain structural flexibility while enduring the load of the drive motor, striking a balance between mechanical stability and functional tactile feedback.

The mechanical properties of TPU were measured as follows: a tensile strength of 29 MPa, an elongation at break of 330% (per ASTM D638 [36]), and a heat deflection temperature of 138 °C. Compared to PLA, TPU demonstrated superior performance. As indicated in

Table 1. Comparison of mechanical properties between PLA and TPU materials.

Property	PLA	TPU	Improvement
Impact Absorption Energy (J/m2)	18 J/m2	54 J/m2	300%
Fatigue Life (cycles)	$2.1\times {10}^5$	$9.8\times {10}^5$	466%
Surface Hardness (Shore)	D75	A95	Improved similarity to skin tactile feel

, TPU provides approximately 300% greater impact absorption and a fatigue life nearly 4.6 times longer than PLA. Moreover, its surface hardness exhibits surface compliance that is perceived by users as more comfortable during physical interaction. These comparative findings validate the mechanical suitability of the selected material for the structural design of robotic hands and establish a critical foundation for delivering tactile realism and emotional engagement in HRI applications.

This study builds upon an open-source model provided by 3DPrintBunny (https://www.thingiverse.com/3dprintbunny/designs (accessed on 3 January 2026)). The original design was reconstructed using Fusion 360 (Autodesk Inc., San Francisco, CA, USA) in alignment with biomechanical standards. Modifications were applied to finger length, joint spacing, and palm thickness to reflect the average proportions of an adult hand, particularly the ratio of distal interphalangeal (DIP), proximal interphalangeal (PIP), and metacarpophalangeal (MCP) joints at 2:3:5. Figure 1 illustrates the CAD layout of the model components (a) and the assembled robotic hand (b). In its initial configuration, the model required five servo motors and a total of 15 wires, encompassing both signal and power lines. This complexity presented challenges in wiring and assembly, identified as significant limitations of the original setup. To address these issues, the Dynamixel AX-12A motor (ROBOTIS Inc., Seoul, Republic of Korea)—featuring UART-based communication—was introduced. As depicted in Figure 2, the AX-12A provides a rated torque of 15.3 kg·cm at 12 V and offers superior precision and synchronization compared to the previously employed PDI-HV2060MG servo motor (Guangdong Chaoli Motor Co., Ltd., Shantou, China). Moreover, AX-12A motors allow for serial communication using only three shared wires, substantially simplifying the overall wiring configuration. To accommodate these motors, the original side-mounted layout was re-engineered into a front-mounted configuration, with redesigned internal forearm geometry and cable routing. An additional MG90S servo motor (TowerPro, Taipei, Taiwan) was also integrated for independent control of the thumb joint, utilizing its own three-wire setup. Through these modifications, the total number of required wires was reduced from 15 to 6, significantly streamlining mechanical wiring and improving system maintainability.

The robotic hands were fabricated using an FDM 3D printer with PLA and TPU materials, specifically tailored for HRI experiments. To enhance tactile sensation during physical interaction, a grid fill pattern was applied to the internal palm structure, featuring an infill density of 10% and a wall thickness of 1.2 mm. This configuration was designed to optimize both structural durability and tactile comfort. The design parameters were informed by the guidelines proposed by Jung et al. [37], which suggest that a 10 to 20% infill density offers an optimal balance between flexibility and mechanical stability in printed structures based on TPU. Figure 3 presents a slicing preview of the robotic hand components printed with FDM. The left image displays the internal infill configuration of the TPU palm structure, while the right image illustrates the layout of the PLA-based finger modules.

Printing Parameters. In this study, appropriate 3D printing parameters were established for PLA and TPU materials based on their distinct thermal and mechanical characteristics. For PLA, optimal settings included a nozzle temperature of 245 °C, a bed temperature of 55 °C, and a print speed of 50 mm/s. These parameters were selected to achieve a balance between interlayer adhesion and surface quality. TPU, a viscoelastic material with high viscosity and low shear strength, is prone to printing issues such as filament stringing and layer separation. To improve the quality of the print, the print speed was reduced to 30 mm/s and the nozzle and bed temperatures were set to 230 °C and 65 °C, respectively. In addition, a retract distance of 1.2 mm and a cooling fan speed of 30% were applied to maintain consistent interlayer bonding. Both materials were printed with a layer height of 0.1 mm to ensure dimensional accuracy and surface uniformity. These parameters were defined according to the material-specific additive manufacturing guidelines proposed by Singh et al. [38], with the aim of optimizing both the stability of the print and the mechanical performance [39,40,41,42].

Table 2. Comparison of FDM 3D printing parameters for PLA and TPU materials, along with recommended ranges for TPU settings based on Singh et al. [38].

Parameter	PLA Setting	TPU Setting	Recommended Range
Nozzle Temperature (°C)	245 °C	230 °C	225–250
Bed Temperature (°C)	55 °C	65 °C	–
Print Speed (mm/s)	50	30	15–40
Retraction Distance (mm)	1	1.2	1–2
Layer Height (mm)	0.1	0.1	≤0.2

summarizes the FDM printing conditions for PLA and TPU, along with the parameter ranges recommended for TPU reported in previous studies.

3.2. Hardware Configuration and Control Architecture

In this study, a power control circuit was designed to incorporate a power management system and a communication interface for operating a robotic hand. A dedicated printed circuit board (PCB) was fabricated to support this configuration. The controller functions as a standalone module, allowing easy attachment and detachment from existing humanoid robot platforms. It was integrated with the upper level control system, ALICE 3 [43], and configured for real-time communication using the ROS-based rosserial protocol. The control architecture is based on an ATmega2560 microcontroller (Microchip Technology, Chandler, AZ, USA). Each finger joint is actuated by four Dynamixel servo motors (AX-12A) and one servo motor (MG90S), which allows flexion and extension movements. The AX-12A motors are precisely controlled via UART signals, achieving a resolution of approximately 0.3°. An additional degree of freedom for the thumb is provided by a servo motor operated via PWM at 50 Hz, offering a control resolution of approximately 1°.

3.3. System Integration and Real-Time ROS Communication

Figure 4 illustrates the overall hardware architecture of the robotic hand control system. The system is powered by an 11.1 V lithium polymer battery (Li-Po), while an R78B 5.0 V DC-DC converter (RECOM Power GmbH, Gmunden, Austria) supplies stable voltage to the MG90S servo motor and the HC-06 Bluetooth communication module (Guangzhou HC information Technology Co., Ltd., Guangzhou, China). The Arduino Mega interfaces with the AX-12A motors, the MG90S servo motor, and the HC-06 module via UART and PWM protocols, and connects to the upper level controller (NUC) through a USB interface. Bidirectional communication with the AX-12A motors is enabled by a 74LVC2G241 level shifter (Nexperia, Nijmegen, The Netherlands) on the data line. Power stability is further ensured by integrating overcurrent protection circuits and decoupling capacitors.

The controller, power distribution circuitry, and communication modules are consolidated on a single PCB platform, allowing independent operation without external control hardware. The PCB is designed with modularity and scalability in mind, which supports seamless integration with other robotic systems. For real-time bidirectional communication and remote control with ALICE 3, the system uses the ROS-based rosserial protocol, which facilitates message exchange, service invocation and timer-based event handling between ROS nodes. All functionalities of the robotic hand controller are managed through ROS topics, services, and timers, ensuring reliable and responsive execution of commands from upper-level nodes.

3.4. Gesture-Based Repetition Test and Joint Angle Analysis

In this study, a repetition test of gesture motions was conducted to evaluate the control reliability and motion reproducibility of a robotic hand in a HRI scenario. The control algorithm was implemented using a ROS-based finite state machine that employed service callbacks and timer events for state transitions. Five gesture states were predefined: Scissors, Rock, Paper, Handshake, and Static Pose. For the experiment, three gestures were selected for testing: Paper (Open), Partial Flexion, and Rock (Closed).

Figure 5 illustrates the repeated gesture motion experiments and the corresponding joint angle measurement method. Each gesture was executed according to a predefined motion sequence designed to ensure high repeatability and structural stability across multiple trials. The effectiveness of these characteristics was quantitatively evaluated using joint angle repetition tests, as shown in

Table 3. Average angle error for each gesture motion.

Gesture	Mean Angle (°)	Standard Error (°)	Standard Deviation (°)
Paper (Open)	1.45	0.078	0.25
Scissors (Partial Flexion)	38.85	0.085	0.27
Rock (Closed)	130.13	0.144	0.46

. Measurements were taken at the Proximal Interphalangeal (PIP) joint of the index finger, with a normal vector perpendicular to the back of the hand used as the reference axis for calculating bending angles. For each gesture, the mean angle, standard deviation, and standard error were computed to assess the consistency and reliability of repeated motions.

The proposed robotic hand is optimized not for high-precision absolute positioning, but for producing consistent and expressive motions suited to HRI applications. Accordingly, the primary performance criterion is not the absolute accuracy of each gesture, but rather the system’s ability to reproduce the same motion reliably in response to repeated commands.

• The Paper (Open) gesture demonstrated the lowest error, with a mean angle of 1.45°, a standard deviation of 0.25°, and a standard error of 0.078°. This high precision is attributed to minimal structural constraints and reduced friction.
• The Scissors (Partial Flexion) gesture maintained a high level of consistency, despite its intermediate flexed configuration.
• The Rock (Closed) gesture, which involves multiple joint movements, exhibited acceptable precision within the expected error margin.

These findings confirm that the TPU-based robotic hand is capable of executing reliable and repeatable motions, even when evaluated solely at the Proximal Interphalangeal (PIP) joint. This result supports its applicability in human–robot interaction scenarios, including emotional expression, object handover, and gesture-based communication.

3.5. Custom PCB Design for Integrated Robotic Hand Control

In this study, a hardware platform for a robotic hand control system was developed using an ATmega2560-based Arduino Mega board (Arduino, Monza, Italy), optimized for laboratory experimentation and testing. To improve system integration and minimize wiring complexity, a custom-designed PCB was fabricated. This PCB consolidates control pins for the servo motors, communication lines for the HC-06 Bluetooth module, and power input and distribution circuitry into a single board, thereby streamlining the hardware configuration. It is engineered to support stable operation of essential components, including the MG90S servo motors (left and right), AX-12A motors (left and right), Bluetooth module, external battery connector, and UART communication ports. The signal routing and layout were carefully designed to facilitate simplified control and enhance overall system reliability. The board features a two-layer structure (top and bottom) with wide traces for high-current power lines, reducing voltage drops and minimizing heat generation during operation. Moreover, the PCB is fully compatible with the Arduino Mega pin configuration, allowing direct mounting without additional wiring. Its modular architecture also enables future upgrades by permitting the microcontroller to be embedded directly onto the board, thereby enhancing compactness and integration for advanced applications.

4. Evaluation

4.1. Experimental Design

This study employed a repeated-measures design in which each participant sequentially interacted with robotic hands fabricated from PLA and TPU materials. To enable within-subject comparisons, a cross-sectional pre-post experimental protocol was implemented. Participants interacted with both versions of the robotic hand, each mounted on a humanoid robot, and performed a series of predefined gestures, including handshakes and grasping tasks (Figure 6).

Participant Design. Participants were recruited from undergraduate students and members of the general public affiliated with Dong-A University, ranging in age from 19 to 65 years. A total of 48 individuals took part in the experiment; however, one response was excluded due to insufficient engagement, resulting in a final sample size of 47. Individuals were excluded if they were unable to read or complete the questionnaire. The inclusion and exclusion criteria, as well as participant demographics, are presented in

Table 4. Participant inclusion and exclusion criteria.

Condition	Description
Inclusion Criteria	Undergraduate students and members of the general public affiliated with Dong-A University, aged between 19 and 65
Total Participants	48 recruited → 1 excluded due to invalid response → Final analysis with 47 participants
Exclusion Criteria	Participants unable to complete the questionnaire

.

Experimental Procedure. As shown in Figure 7, prior to participating in the study, all individuals were informed of its purpose and procedures by the researcher and provided written informed consent. During the first session, participants interacted with a robotic hand fabricated from PLA material, performing a series of predefined gestures including a handshake, rock-paper-scissors, and a neutral pose. Immediately following this interaction, they completed a post-experience questionnaire. After a 5-min rest period, the second session commenced, in which participants engaged with a TPU-based robotic hand. The same set of tasks was performed, followed by completion of the identical questionnaire. Upon completing both sessions, participants were asked to indicate their preferred robotic hand (PLA or TPU) and to provide open-ended feedback explaining their choice and offering any additional comments.

Equipment and Materials. The questionnaire utilized in this study was developed based on Dautenhahn’s HRI evaluation framework [29]. It incorporated items rated on a 5-point Likert scale to assess participants’ perceptions across six core dimensions: motion consistency, grip force, response speed, satisfaction with surface texture, satisfaction with hardness, and emotional response. A summary of the evaluation dimensions, along with representative questionnaire items, is presented in

Table 5. Evaluation items and corresponding questionnaire focus.

Evaluation Category	Key Questionnaire Items
Motion Consistency	Predictability of movement, accuracy in object manipulation
Grip Force	Appropriateness of force control, satisfaction with grip strength
Response Speed	Satisfaction with the robot hand’s response time
Surface Texture Satisfaction	Perception of the hand’s tactile feel and texture
Hardness Satisfaction	Subjective evaluation of the hand’s firmness
Emotional Response & Preference	Overall emotional reaction, subjective comfort, material preference

.

4.2. Data Collection and Analysis Methods

Data for this study were collected via an online questionnaire administered through Google Forms. Following interaction with each version of the robotic hand, participants completed the survey, and their responses were automatically recorded in Google Sheets. Responses to open-ended questions were extracted independently and converted into plain-text format for subsequent qualitative analysis. The questionnaires were designed in accordance with the HRI evaluation framework and distributed digitally via Google Forms, as illustrated in Figure 8.

Quantitative Analysis. Descriptive statistics, including means and standard deviations, were computed for all quantitative measures. To compare satisfaction ratings between the two conditions (PLA hand vs. TPU hand), paired t-tests were performed, accounting for the within-subject structure of the repeated-measures design. Data processing and statistical analysis were conducted using the Python 3.8 (Python Software Foundation, Wilmington, DE, USA) programming language, specifically leveraging the pandas, numpy, and scipy.stats libraries. Statistical significance was evaluated using a threshold of p < 0.05.

The analysis focused on five evaluation categories:

• Motion consistency
• Response speed
• Grip force satisfaction
• Surface texture satisfaction
• Hardness satisfaction

Mean scores were calculated for each material condition (PLA and TPU), and paired t-tests were applied to evaluate differences across conditions.

Qualitative Analysis. Responses to open-ended questions were analyzed using a content analysis approach. Emotional responses were classified into major affective categories, including comfort, novelty, and discomfort. The frequency of each category was calculated to support comparison and interpretation alongside the quantitative findings. The questionnaire comprised 21 items, of which 9 (Q1, Q2, Q3, Q12, Q13, Q14, Q15, Q16 and Q17) were directly aligned with quantitative evaluation metrics—namely, motion consistency, grip force, response speed, surface texture satisfaction, and hardness satisfaction in

Table 6. Questionnaire items used in the quantitative evaluation, grouped by five HRI-related categories and corresponding question numbers.

Evaluation Category	Question No.	Summary of Questionnaire Item
2. Grip force satisfaction	3	Did you feel the grip force was appropriately controlled?
	12	Was the level of pressure appropriate?
	15	Was the grip strength sufficient?
3. Response speed	17	How would you rate the robot hand’s response speed?
4. Surface texture satisfaction	13	Were you satisfied with the surface texture of the robot hand?
5. Hardness satisfaction	14	Was the hardness of the robot hand at an appropriate level?

. The remaining items addressed emotional reactions, affective impressions, perceived safety, overall satisfaction, and user suggestions, serving as the basis for qualitative and supplementary analysis. A summary of these items and their respective analytical objectives is provided in

Table 7. Summary of questionnaire items used for qualitative and supplementary analysis, including emotional and subjective response categories.

Item No.	Summary of Question	Purpose/Reason for Inclusion
4	Did the robot hand’s tactile sensation feel natural?	Assesses tactile realism as a separate emotional perception from overall surface satisfaction.
5	Was it comfortable to interact with the robot hand?	Emotional/affective response item for separate qualitative analysis.
6	Did you feel safe while using it?	Used for qualitative assessment of perceived safety and trust.
7	Did you enjoy the interaction?	Assesses emotional engagement and positive affect.
8	Did the interaction feel natural?	Evaluates overall fluency of the interaction; supports synthesis of qualitative impressions.
9	Overall satisfaction	Broad summary item; used as reference rather than core metric.
10	Did the tactile sensation feel natural?	Overlaps with Item 4; may be used separately or omitted due to redundancy.
11	Did it feel similar to a human hand?	Evaluates HRI-related subjective perception; part of qualitative assessment.
18	What emotions did you experience?	Open-ended emotional response; not suitable for quantitative analysis.
19	Did you feel uncomfortable or anxious?	Used as an indicator of negative emotion or psychological discomfort.
20	Suggestions for additional features	Open-ended input; used for exploratory analysis.
21	Any other suggestions or comments	General comments section; used for qualitative content analysis.

.

4.3. Experimental Results

This study involved 47 participants who interacted with robotic hands made from PLA and TPU materials. After each interaction session, the participants completed a structured questionnaire to evaluate their experiences. The responses were analyzed to compare user perceptions between the two material conditions.

Quantitative Results. User satisfaction was assessed across five key dimensions: motion consistency, grip force, response speed, surface texture, and hardness. In all categories, the TPU-based hand received higher average ratings than the PLA-based hand, with statistically significant differences observed. The TPU hand consistently outperformed the PLA hand in terms of average scores across all five dimensions. In particular, satisfaction with surface texture averaged 4.23 for the TPU hand, compared to 2.87 for the PLA hand. Additionally, hardness satisfaction averaged 4.40 for the TPU hand, while the PLA hand also scored 2.87. These results provide empirical evidence that TPU offers a more natural and emotionally satisfying tactile experience in human–robot interaction contexts. A detailed comparison of the average scores for each evaluation item can be found in

Table 8. Comparison of mean satisfaction scores between the PLA and TPU robotic hands across five evaluation categories (5-point Likert scale).

Evaluation Category	PLA Mean (5-Point Scale)	TPU Mean (5-Point Scale)
Motion consistency	3.66	4.14
Grip force	3.79	4.38
Response speed	3.66	3.96
Surface texture satisfaction	2.87	4.23
Hardness satisfaction	2.87	4.40

. Since the same participants interacted with both the PLA and TPU versions of the robotic hand, we conducted paired t-tests to determine if there were statistically significant differences in user perceptions. As summarized in

Table 9. Paired t-test results comparing user satisfaction between PLA and TPU robotic hands across five evaluation categories.

Evaluation Category	PLA Mean	TPU Mean	t-Value	p-Value	Sig.
Motion consistency	3.66	4.14	−3.453	0.0012	Yes
Grip force	3.79	4.38	−5.529	<0.001	Yes
Response speed	3.66	3.96	−2.455	0.0179	Yes
Surface texture	2.87	4.23	−9.672	<0.001	Yes
Hardness satisfaction	2.87	4.40	−10.313	<0.001	Yes

, all five evaluation categories showed significant differences ($p<0.05$). Surface texture satisfaction ($t=-9.672$, $p<0.001$) and hardness satisfaction ($t=-10.313$, $p<0.001$) exhibited the strongest effects, indicating that the TPU material provided a superior tactile experience compared to PLA.

Interestingly, while PLA is physically harder than TPU, participants rated the TPU hand significantly higher in terms of perceived hardness satisfaction. This indicates that the participants judged the firmness based on their personal comfort rather than on objective measurements of stiffness. The soft and compliant surface of the TPU hand likely contributed to a more natural and emotionally comfortable interaction. These findings support the conclusion that user perception in HRI is influenced not only by mechanical performance but also by factors such as emotional comfort, tactile realism, and perceived naturalness.

Qualitative Results. Participants were asked an open-ended question: “What kind of emotions did you experience during your interaction with the robotic hand?” Content analysis revealed several dominant emotional categories, including curiosity/novelty, comfort, enjoyment, and neutrality. In general, the TPU-based robotic hand elicited stronger emotional responses compared to the PLA hand, ranging from curiosity to comfort and immersion. In contrast, the PLA hand was often associated with feelings of stiffness and mechanical detachment. These findings suggest that the tactile properties of robotic hands can significantly influence user emotional perceptions and overall quality of interaction (see

Table 10. Representative participant responses categorized by emotional type for PLA and TPU robotic hands.

Emotion Type	PLA Example Response	TPU Example Response
Curiosity/Novelty	“It felt mechanical but interesting.” “The way the robot hand moved was strange but fun.”	“It felt like a real human hand, so it was fascinating.” “I was surprised by the natural movement.”
Discomfort/Alienation	“It felt stiff and uncomfortable.”	(Almost no responses)
Comfort	“It felt hard, but not as bad as I expected.”	“The touch was soft and comfortable.” “The gentle grasp felt just like a real human hand.”
Enjoyment/Fun	“Playing rock-paper-scissors was fun.” “Interacting with the robot hand was interesting.”	“The hand-grasping motion was fun and cute.” “It responded well, so I got immersed.”
Indifference/Neutral	“I just thought of it as a robot.” “I didn’t feel any particular emotion.”	“It didn’t feel very different, but it was natural.”

).

Preference Analysis. When asked to choose their preferred robotic hand, 29 out of 47 participants (61.7%) selected the hand based on TPU. In contrast, only six participants (12.8%) preferred the PLA version, while 12 participants (25.5%) had no clear preference. The strong preference for the TPU hand aligns with both quantitative and qualitative findings, especially in terms of tactile satisfaction, natural movement, and emotional acceptance. These results suggest that users not only recognized the differences but also actively favored the enhanced tactile and emotional qualities of the TPU-based robotic hand.

4.4. Experimental Discussion

This study aimed to empirically assess how material properties influence user perception in HRI. We evaluated two identically structured robotic hands made from different 3D printing materials: TPU and PLA. The quantitative results clearly showed that the TPU-based hand consistently outperformed the PLA hand in the five evaluation categories: motion consistency, grip force, response speed, surface texture satisfaction, and hardness satisfaction (see Table 8). Paired t-tests confirmed that these differences were statistically significant. In particular, strong effects were observed in categories related to tactile experiences, such as surface texture ($t=-9.672$, $p<0.001$) and perceived hardness ($t=-10.313$, $p<0.001$) (see Table 9). In particular, while PLA is physically harder than TPU, participants rated their satisfaction with the TPU hand higher in terms of perceived hardness. This suggests that the evaluation of “hardness” was not solely based on the actual stiffness of the material, but rather on how appropriate the firmness felt in a human–robot interaction context. The soft and more flexible nature of the TPU hand likely contributed to a more psychologically comfortable and emotionally acceptable tactile experience. Qualitative feedback from participants supports this interpretation. Descriptions of the TPU hand included terms such as “soft,” “comfortable,” “natural,” and “similar to a human hand.” In contrast, participants described the PLA hand with phrases like “stiff,” “mechanical,” and “cold.” These emotional impressions align with the quantitative findings, confirming that the material properties of the hand, specifically its softness and texture, play a crucial role in shaping users’ emotional responses and overall experiences. Furthermore, the user preference survey revealed that 61. 7% of the participants preferred the TPU hand over the PLA hand. This reinforces the idea that materials that enable more natural and emotionally rich interactions are more desirable in human–robot interaction contexts.

This study presents empirical evidence on how material properties affect the quality of HRI. It isolates material as the only independent variable while keeping the hardware structure identical. By controlling all other design factors, this research ensures that any observed differences in user responses are solely due to variations in material. This rigorous approach improves our understanding of how physical materials influence tactile perception and emotional responses. The findings highlight that the choice of materials goes beyond just mechanical performance or ease of fabrication; it significantly influences the trust of the user, the emotional comfort, and the naturalness of physical interactions. The excellent performance of TPU in both quantitative and qualitative assessments underscores its potential as a preferred material for creating robotic hands in emotionally sensitive applications. These findings have significant implications for the design of robots that interact closely with humans, including social robots [44], caregiving robots, and educational platforms. Designers and engineers are encouraged to integrate emotionally acceptable tactile features, such as compliance, softness, and skin-like textures, into the development of HRI systems. This focus on material characteristics represents a move towards more human-centered robotics, where the emotional quality is treated as an essential functional requirement rather than just a secondary aesthetic consideration.

In the field of modeling-driven soft gripper design, our findings on perceived comfort can be viewed as a valuable, user-centered design insight. Wang et al. [45] showed that prestressed soft grippers can be accurately modeled and manufactured to achieve predictable bending behavior and stable grasping during food handling tasks. Similarly, Bo et al. [46] introduced a data-driven topology optimization framework that uses grasp simulation forces as criteria for optimizing gripper components. Although our study does not involve finite element modeling (FEM) or mechanical testing, the comfort-driven structural patterns identified in our user study could be translated into target ranges for contact area, normal force distribution, and required stiffness. These factors could then inform modeling and topology optimization processes. In this way, our work connects subjective preferences with quantitative design requirements, indicating that “perceived comfort” should be considered an additional design objective alongside grasp success rates and material efficiency in future modeling-driven soft gripper design.

This study has several limitations despite its contributions. First, the experimental design was limited to short-term interaction sessions, which means that it does not account for long-term usability, the development of trust, or habituation effects that may occur with repeated interactions with the robotic hand. Future research should include longitudinal studies to better understand how user perceptions evolve over time. Second, the participant pool was made up only of adults aged 19 to 65 years, which may restrict the generalizability of the findings to younger or older populations, such as children or the elderly. Different age groups may perceive tactile and emotional signals differently, highlighting the need for a larger demographic sampling in future studies. Third, the interaction tasks were confined to basic gestures such as handshake, rock-paper scissors, and a neutral pose. Although these tasks are sufficient to assess core tactile attributes, they may not fully capture the richness and complexity of real-world HRI scenarios. Additional interaction contexts, including object manipulation, social touch gestures, or cooperative tasks, should be explored to enhance the applicability of the results. Lastly, although the study focused on material-driven differences in HRI, other factors such as visual appearance, motion smoothness, and voice feedback were kept constant. Future studies could investigate how these additional modalities interact with tactile experiences to influence emotional acceptance and user satisfaction.

5. Conclusions and Future Work

This study provides empirical evidence that the material composition of 3D-printed robotic hands significantly influences tactile perception and emotional acceptance in HRI. By comparing anatomically identical hands made from PLA and TPU, the experiment confirmed that TPU yields superior perceptual performance, particularly in surface texture satisfaction and perceived hardness. Participants consistently reported greater emotional comfort and realism with the TPU-based hand, supporting its applicability in caregiving, educational, and therapeutic settings. The findings emphasize that tactile properties are not merely aesthetic considerations but functionally tied to emotional receptivity and user acceptance, key factors in the design of socially engaging robots. Although the results are statistically significant, limitations such as material scope, participant diversity, and duration of the interaction suggest directions for improvement.

Future research will expand material comparisons to include silicone-based elastomers and conductive polymers, incorporate multimodal feedback systems (e.g., haptic, auditory, and visual cues), and diversify participant demographics. Furthermore, employing validated affective measurement tools, such as the Self-Assessment Manikin (SAM), may enable deeper insight into emotional responses during human–robot tactile interaction. This study lays the foundation for the development of emotionally intelligent robotic systems that foster trust and engagement through material-driven design.