Touch care robots, designed to provide 'tactile touch' and embody the concept of 'humanitude' are a burgeoning trend in contemporary human-robot interaction, particularly in medical settings. Prior studies have indicated that several factors can impact a robot's tactile perception. Within these environments, fully human avatars exhibit a resemblance to robots and can evoke a sensation akin to interacting with a human. To explore this relationship, we conducted a user study with 26 participants, focusing on the impact of avatars with distinct visual characteristics (muscular vs. skinny, male vs. female appearance) on the robot's tactile perception. The results revealed disparities in the influence of visual appearance on haptic perception, as well as gender-based variations in preferences. Additionally, the gender of the user constitutes a crucial factor in the perception of user-robot interaction and warrants consideration.
ing placed on the development of robots that can replace people in providing such touch care.
It has been demonstrated that stroking movements have a relaxing efect on the body and mind by mediating the autonomic nervous system and the endocrine system, and 2. Related Work there is a technique called touch care, which is medical care that incorporates stroking movements. In general, A number of relevant studies on robot touch have shown touch care refers to all care that includes stroking move- that robots induce positive emotions in humans through ments and has various efects, such as reducing anxiety touch interactions [1, 2]. The basis for this efect is a muland stress and relieving pain. In addition, the stroking timodal care approach based on philosophical principles movements of touch care are also incorporated into care known as ’tactile touch’ and ’human touch’. They are such as tactile massage and humanitude care, which are practiced by humans in nursing and other medical fields. practiced in the medical field. Thus, touch care is recog- While these methods are practiced by humans in everynized as a method suitable for reducing anxiety and pain day situations, problems related to human resources and in actual medical and nursing care settings, is an essential nursing infrastructure remain [3, 4]. Therefore, recent type of care, and is expected to remain in high demand. research has focused on the development of touch care However, the recent shortage of human resources due to robots. The aforementioned technological implementaa shortage of carers and "caregiver fatigue" among home tion is an example of an attempt to replicate the practice carers has become a serious problem, and emphasis is be- of reducing patient stress and relieving pain through gentle touch interactions. This touch is influenced by various aspects such as stroke speed and speech [5]. While there have been many mechanical and technical studies on touch care robots, few studies have focused on the visual representation of these robots and their impact on patients. Going further into the actual touch care robots, the gender of the avatar as visualized in an augmented reality (AR) environment afected the overall liking [ 6, 7].
an Alienware x17 laptop with a built-in NVIDIA GeForce 3080ti. Moreover, Unity 2021.3.13f1 was used for the experimental setup. For the gentle touch, a Universal Robots UR3e with 3D printed end efector is used. The right controller of the Oculus Quest 2 to the last joint of the robot arm to enable hand tracking. Furthermore, All questionnaires except a demographic questionnaire and a final feedback questionnaire were included in the virtual contents. The participant could use a controller to answer the questionnaires inside virtual environment, but we also provided alternative questionnaires outside of virtual reality in case there were any problems. The touch was performed by the UR3e robot and by tracking the mounted controller synchronized with the avatar’s hand movement using Final IK VR which is a high-speed full-body solver dedicated to animating VR avatars. For the touch, the program on the UR3e robot was set to slowly press down until a pressure of 3 [N] was detected. Once this is done, the pressure is held for 3 [sec] and then the arm swipes two times along the participant’s arm and back. It then returns to a zero position where it no longer touches the participant’s hand. In the virtual environment, the participant could see the avatar’s arm move according to the robot’s arm in the real world (See Fig. 1).
avatars and two skinny (male and female) avatars. The animations were done using Maximo, and the audio voices were using Narkeet. The avatars were created using DAZ Studio according to Kocur et. al [8]. The used avatars can be seen in Fig. 2.
characteristics of avatars impact both the perception of the touch care robot itself and the perception of touch.
For this study, we employed the Godspeed Like-ability Scale along with adjective pairs to assess touch perception. The experiment followed a specific procedure. First, written informed consent and a brief overview of the study procedure were given to the participant. After signing the informed consent form, the participant completed a demographic questionnaire along with some additional questions about visual impairment and cultural background. Once completed, the participant was seated, the Oculus Quest 2 (VR-HMD) was fitted, and the participant was given the left-hand controller. The participant was instructed to place his or her left arm on the cushion in front of two pads once the avatar in the virtual environment had finished introducing himself or herself. When the participant entered the virtual environment, the first avatar walked toward him/her and introduced himself/herself. After placing the hand correctly, the instructor started the robotic touch procedure.
The visualization of the touch is seen by the participant in VR as seen in Fig. 1. Following the robotic touch, the avatar left the scene, and the participant regained the ability to move their arm. Subsequently, the participant proceeded to complete the Godspeed Like Ability scale, which assessed their overall perception of the robot and the experience. Upon completing the questionnaire, the next avatar entered the scene. Four distinct avatars were utilized, and their usage was balanced through the implementation of Latin squares. Instead of walking in, the Figure 3: Evaluation of the Godspeed series likeability within avatars transitioned directly from a resting state to the Muscular avatar (blue graph) and Skinny avatar (orange touch sequence. Following that, the participant was re- graph). quired to answer the questionnaires. Upon concluding the study, participants were provided with a feedback form to gather their subjective impressions regarding their perception of the robot and the robotic touch. 4.2. Participants context of robot touches by Yamashita et al., where participants were asked to rate a set of opposing adjectives [13]. Similarly, the same pair of adjectives was employed by other researchers, such as Endo et al., in evaluating their robotic hand [14]. This particular set of adjectives appears suitable for conducting a comprehensive investigation into the variation of individual properties across diferent stereotypes.
We recruited 26 students (21 male, 5 female) between 23
and 37 years old (M = 28.35, SD = 4.65) from our
university. We determined the sample size using a prior
power analysis and used G*Power (Version. 3.1.9.7) and
assumed to conduct a two-way ANOVA with a factor of
three levels with an level of 0.05, power of 0.80, corre- 5. Result
lation of 0.5, of 1, and efect size f of 0.25 (medium). The
results indicated that 24 participants would be required. From the result no efect on anything regarding
GodThe research was conducted in a separate room within speed Likeability Scale. We evaluated each questionnaire
the university building. Both demographic and feedback using two-way repeated measure ANOVA models. We
questionnaires were administered anonymously. All par- also calculated linear models. The analysis was done
ticipants received a gift card as a token of appreciation using Python together with the Pandas, Seaborn, and
for their participation. This study was approved by the SciPy packages. The Godspeed questionnaire as well
Ethics Committee of the Nara Institute of Science and as the opposite adjective parts were also evaluated. We
Technology and was conducted according to institutional also performed evaluations by gender of the participants.
ethical provisions and the Declaration of Helsinki. Looking at the result, there was no significant efect of
F(1) = 0.673 (p = .414), GENDER F(1) = 3.220 (p = .076).
4.3. Evaluation method An overview of the efects can be seen in Fig. 3 and
Fig. 4. Some efects on the touch adjectives: For all
Employing a variety of questionnaires is a widely adopted participants, dangerous vs safe (Skinny significant more
method for assessing human-robot interaction (HRI) sce- safe than muscular and female significant more safe than
narios. The Godspeed series is a renowned questionnaire male), dislike vs. like (Female significant more like than
that ofers tools for evaluating robot anthropomorphism, male), frightening vs. gentle (Skinny significant more
animacy, likeability, perceived intelligence, and perceived gentle than muscular), negative sensation vs positive
sensafety [9]. Each of these tools can be independently uti- sation (Female significant more positive sensation then
lized, for example, the likeability questionnaire has been male), soft vs hard (Male significant more hard then
feemployed in various HRI studies [
There was also a significant efect on GENDER F(1) = 6.596 (p = .012) in the second part regarding undesirable vs. desirable. The same was found on GENDER F(1) = 7. Conclusion 4.686 (p = .033) for negative vs. positive sensation. We also found a significant efect of GENDER F(1) = 5.775 (p = .018) in the second part for negative vs. positive feelings.
This study investigates how users perceive touch care robots with diferent avatars, considering specific visual features’ influence on touch and overall perception. 26 participants explored how avatars with varying visual features afected their tactile perception of the robot. 6. Discussion Results showed variations in the impact of visual appearance on haptic perception, along with gender-based In the study, we found that the Skinny avatar has a safer preference diferences. Comparing muscular and skinny and more gentle impression than the muscular one, and avatars, the skinny one created a safer and gentler impresalso focuses on gender, the female avatar has more safe, sion. Thus, the appearance of touch care robots in virmore like, more positive, more pleasant, and more desir- tual reality significantly influences user-robot interaction able for all participants. An avatar based on although perception. The limited number of female participants the GENDER did not play a role for the same ability, the compared to males is a study limitation. Future studmuscular and thin avatars influenced the perception of ies should consider cultural, educational, and religious the touch by the GENDER of the avatars in some cases. perspectives’ influence on perceptions. This work was supported by JSPS KAKENHI Grant Numbers JP19H01124 and JP23H03442.