Keywords

1 Introduction

Japan is known as the first country that became a hyper-aged society. The percent of people aged 65 year and above in Japan has exceeded 25% in 2013. Friedman et al. described one of the key factors to successful aging as being the maintenance of an active relationship with the surrounding community [1]. During the 1970s and 1980s, many new towns were constructed in the suburbs and the aging rate of the population in these districts is especially rapid. Many of these areas have an elderly population already above 30% and they are going to become marginal villages in near future. The major reason is that younger generations in those areas are likely to move to urban areas when they grow up. Though the elderly living in urban district can still benefit various kinds of relationships in the local communities, the elderly living in suburban districts are facing the risk of losing active relationships in local communities. Information technology (IT) is considered a powerful tool to maintain a social presence for the elderly [2]. Comparison between an older Internet user and an older non-user showed that Internet use has a positive effect on the user’s psychological well-being [3]. Our research group aims to develop a model wherein local senior communities support each other by connecting those communities through the Internet.

Takagi et al. proposed a system that connects an urban senior community and a suburban senior community using video conferencing to offer a lifelong learning service—IT classrooms—from urban area to suburban area. IT classrooms are mainly located in certain urban areas. Therefore, senior communities in suburban areas, like the new towns mentioned earlier, have difficulty accessing such classes. In contrast, many older adult IT experts do not have opportunities to use their skills in their own communities [4]. From this study, participants mentioned that it would be helpful if it were possible to convey and receive non-verbal information to remote participants and to grab their attention.

One possible solution to this issue is the application of a telepresence robot. A telepresence robot is the strongest media that enables embodied interaction with a distant person. Application of telepresence robots has begun and has made it possible to expand the interaction of the elderly with family or close friends. Many kinds of telepresence robots have become commercially available and are easy to develop from off-the-shelf robot platforms [5]. Remote elderly care is expected to be one of the most promising applications of service robots [6]. The ExCITE project focused on designing a social telepresence system for elderly care using the mobile telepresence robot Giraff [7]. Many studies have been done that examine the effect of telepresence systems in education or communication environments with younger participants. Tanaka et al. compared the difference in communication using a video conferencing system and a mobile telepresence robot in remote foreign language class for children [8, 9]. In this study, application of the telepresence robot encouraged vocalization by the children. Rae et al. explored how the height of the mobile telepresence robot [10] and the task conducted by the robot [11] affects working efficiency and the form of communication. Social presence of the remote operator can also be improved even if a remote operated system is designed without mobility [12, 13]. Tsui et al. examined cases of actual office meetings and found that the telepresence robot fit more in the situation like informal meetings [14]. From previous research, the telepresence robot will be a useful tool for remote communication. Beer et al. assessed that older adults are eager to control telepresence robot for having communications or experiences in distant location, however, the presence of robot in their private space make them feel awkward [15]. This implies that public space is more acceptable place for installing telepresence robot than private space.

This study aims to explore the feasibility of telepresence robots to facilitate communication between distant senior communities in public situation. In Sect. 2, we describe conditions of the application field. Then, in Sect. 3, we introduce the designed system, which uses telepresence robots for remote IT education between distant senior communities followed by the evaluation and discussion of the system in Sects. 4 and 5.

2 Application Field

We applied the system to a remote IT course between Sendai City and Kiyosedai Town in Nishinomiya City. The locations of each place is shown in Fig. 1. The distance between these two sites is about 625 km. Students in this course are residents of Kiyosedai. Kiyosedai is one of the new towns in Japan, and it has a high ratio, estimated at around 35%, of people over the age of 65. There is no available IT classroom for the elderly in Kiyosedai therefore the residents’ association of Kiyosedai decided join this remote IT education course. Lecturers and assistants are members of Sendai Senior Net Club, an IT training organization for the elderly in Sendai. About 100 members have teaching experience in IT classes. Detailed information about these two sites is described in [4].

Fig. 1.
figure 1

Location of learners’ site Kiyosedai, and lecturers’ site Sendai

Full size image

2.1 Participants

Learners were from Kiyosedai and lecturers and assistants were from Sendai Senior Net Club. Classes were mainly held on Sunday mornings and afternoons. There were 12 people who registered as students (7 women and 5 men in their 60s and 70s) in Kiyosedai. Students were divided into two groups (each group had six students). All students owned a laptop or desktop computer but only one student owned a tablet terminal. There were 2 lecturers (both men in their 60s and 70s) and 7 assistants (4 women and 3 men in their 60s and 70s) in Sendai.

2.2 Procedure

The course used for the evaluation of the proposed system started from August 24th 2013 and ended on October 7th. One course was composed of five lessons. Each lesson began with a greeting from the lecturer and an introduction of the day’s topic followed by the lecturer’s presentation and demonstration of how to use the tablet terminal. After the lecture, students began an exercise involving their tablet. During the exercise, remote assistants monitored the students and the students’ interaction with the tablet through the video streamed by the robot and the gesture visualization system. Assistants would help students if students ask for help or if the students seemed to have trouble using the tablet. Then, if all the students finished the required tasks of the exercise, the lesson moved on to the next topic and the lecturer began instruction again.

2.3 A Known Issue

Typical IT class that teach senior citizens how to use a tablet terminal consist of one lecture and have assistants for each group of 6 students for maximum [4]. How organizers take care of novice senior citizens with different learning speed is an important point for a successful class. In an earlier study, lecturers gave a presentation about the topic covered in a class and assistants observed gestures of the learners on the tablet terminals if they faced any troubles. Once the assistant noticed that a student needed help, the assistant could directly communicate with the student. More detailed descriptions about the configuration of the class and gesture visualization system were published in a previous paper [4]. In these former settings, the assistants mentioned that they could not take care of the students sufficiently, since they felt hard to get students attention to themselves and catch the timing to start conversation. This could have been due to the lack of non-verbal interactions in the video conferencing system. We decided to apply a telepresence robot in order to add physical motion as non-verbal cue to convey remote operators’ intention to start conversation.

3 Robots Used for Remote Communication

3.1 System Requirements

Each class had 1 lecturer and 2 assistants. An assistant took care of a group of students. We needed 3 telepresence robots for a class. The telepresence robot for the lecturer needed to be operated in hands free since the lecturer must give a presentation and demonstration of how to use the tablet terminal. On the other hand, the telepresence robots for assistants needed to be operated manually. Assistants are needed to move the camera directly towards the students’ face. Based on related research, different sizes and the mobility of the robot affect the social presence of the remote operator in a different way [10,11,12,13]. For a comparison, we adopted 2 types of telepresence robots for assistants. One was a mobile type robot that allowed the remote operator to move around the class; the other robot was a desktop type robot that the operator could pan and tilt in the direction of a camera. Figure 2 illustrates the overall system configuration of the system used in the IT classroom.

Fig. 2.
figure 2

System overview

Full size image

3.2 Cogi: Lecturer Telepresence System

Cogi receives commands, including the angles of the arm using serial connection over a Bluetooth from PC or mobile devices. In our study, we have built our own video conferencing system based on Web-RTC with the capability to detect the head position of the user in front of the camera. Cogi is linked to this video conferencing system both as a video streaming source and as a receiver of commands transmitted from a remote site. Cogi has two mode to be linked with the videoconference:

  • “Master” mode, which tracks and follows the head of the operator in front of the camera, and sends the movement of the user to the remote Cogi in “Slave” mode.

  • “Slave” mode, which pans and tilts the camera and screen according to the movement sent from Cogi in “Master” mode.

In our study case, the lecturer joins the videoconference with Cogi linked in “Master” mode. Cogi tracks the head of the lecturer and sends the angles of the lecturer’s face from the screen. In the remote classroom, there is another Cogi joining the videoconference that is linked in “Slave” mode. This Cogi directs its face to the angles sent by the Cogi in “Master” mode. For example, when the lecturer looks into the screen from the left, the Cogi in the remote classroom faces to the right. In this way, the lecturer can look around in the remote classroom just by moving his/her head as if he/she were looking through an observation window [16] (Fig. 3).

Fig. 3.
figure 3

Lecturer telepresence system (Cogi). The picture on the left is the Cogi set to “Slave” mode and the picture on the right is the Cogi set to “Master” mode. The “Slave” mode Cogi is placed at the Student site and the “Master” mode Cogi is placed at the Lecturer’s site.

Full size image

3.3 Mobile Telepresence System

The left part of Fig. 4 is the mobile telepresence robot (Double Robotics Double) we used and the right part of Fig. 4 is the captured desktop screen image of the telepresence robot’s control interface. The control interface is a web browser based system. The interface consists of a video window captured by a robot that is then displayed on a robot user interface (UI). Assistants can move the robot with a cursor key on a keyboard. Remote site view is the video from the wide-angle camera placed in the classroom in the community center. The lecturer’s tablet view is a captured video that displays how the lecturer is operating a tablet terminal. In the lower side of the window, the students’ tablet view is displayed. Gesture interaction of a student is overlaid on the captured screen image of a student’s tablet terminal. In order to avoid cross talk, an assistant uses headset for audio communication. A webcam that captures an assistant’s face is attached to the top of the display.

Fig. 4.
figure 4

Mobile telepresence robot (Double) and its control interface for remote assistant.

Full size image

3.4 Desktop Telepresence System

Figure 5 shows the desktop telepresence robot (Evolve robotics Kubi) we used and the screen image of the robot’s control interface. Only the robot UI view is different from the mobile robot’s interface. Assistants can control the desktop telepresence robot by clicking the cursor button on the screen. In addition, assistants can move the display, mounted on a Kubi, toward each learner’s face or tablet by clicking the buttons placed under the video image of the robot UI. Also, the webcam is attached on top of the display.

Fig. 5.
figure 5

Desktop telepresence robot (Kubi) and its control interface for remote assistant.

Full size image

4 Evaluations

Evaluation of the system was made from participants in 5 lessons (Fig. 6). Each lesson held twice during a course. Students are divided into two groups, which supported by either Kubi or Double. Number of attended students in each lesson was 7 in lesson 1 (3 assisted through Kubi, 4 assisted through Double), 8 in lesson 2 (4 assisted through Kubi, 4 assisted through Double), 7 in lesson 3 (3 assisted by Kubi, 4 assisted by Double), 7 in lesson 4 (4 assisted by Kubi, 3 assisted by Double), and 8 in lesson 5 (4 assisted through Kubi, 4 assisted through Double). After each lesson, we asked all lecturers, assistants, and students to answer the questions. We also observed the recorded videos of the lessons and examined the differences in communication among participants from previous remote IT courses.

Fig. 6.
figure 6

Students’ site with three telepresence robots in Kiyosedai. There is a screen that displays the tablet terminal of the lecturer in front of the students. Three robots are placed in the room as indicated in the picture. There are local supporters, two people standing behind the students, that help students in cases where the remote assistants cannot handle all of the work or if there is audio connection trouble.

Full size image

4.1 Lecturers’ Impressions of a Telepresence Robot

Figure 7 shows the answers to the evaluation questions given by the lecturers. From these results, the lecturers’ appeared to get more accustomed to the system and their impressions improved with each additional lesson.

Fig. 7.
figure 7

The lecturers impression compared with teaching in person. From the beginning they felt that they could teach better if they gave a lecturer in person. As lessons went on, they get used to the system and they thought that they can teach as well as face to face.

Full size image

4.2 Remote Assistants’ Impressions of Telepresence Robots

Figure 8 shows the overall impressions of the class as given by the assistants in Sendai. Impressions of assistants, those who use Double, seem gradually get used to operating a robot. Impression of assistants, those who use Kubi, marked a negative impression on the first and the last lessons. They mentioned that time lag of the Internet connection made them difficult to understand the context of remote site. However, Kubi had a good reputation on its pan and tilt motion, since assistants can observe both students face and hands. They also felt some awkwardness at the beginning of the course and they gradually became accustomed to giving support to the students through the robot.

Fig. 8.
figure 8

Impressions of assistants giving technical support to students through both types of telepresence robot.

Full size image

Results in Fig. 9 show the effect of using the telepresence robots for improving the communication quality with students. By using either type of telepresence robot, assistants were able to understand what was going on with the students from the beginning of the course (see the left of Fig. 9). Assistants also felt at ease talking with students by moving the robots. Kubi can directly point the face of certain students and Double can approach toward the students to get their attention (see the right of Fig. 9).

Fig. 9.
figure 9

Two evaluation questions indicating that the telepresence robots improved understanding at the remote site and helped smooth lecturer and assistant communication with students.

Full size image

4.3 Students’ Impressions of Telepresence Robots

Figure 10 shows the impressions that students developed from Cogi, the telepresence lecturer robot. Students also showed a trend similar to that of the lecturers and the assistants. In the beginning of the course, the students would have preferred a face-to-face type of class. However, the students adjusted to learning from the teleoperated robots as the class continued. There was a correlation between preference and number of lessons (r2 = 0.26, p < 0.01).

Fig. 10.
figure 10

Answers to evaluation questions indicating the students’ impressions toward the lecturer telepresence system, Cogi.

Full size image

Although, the transition of the evaluation is slightly different between Double and Kubi, the results in Fig. 11 illustrate that application of both telepresence robots increased the sense of togetherness for the students.

Fig. 11.
figure 11

Evaluation of togetherness by applying telepresence robots.

Full size image

4.4 Did Telepresence Robots Facilitate Communications?

We compared the remote education with telepresence robots and the previous systems in terms of communication facilitation. In previous systems two different types of remote assistance were conducted. The first type was remote assistance by only observing a gesture visualization system [4]. In this case, the assistants just observed a gesture visualization system and if they found that the students were facing some difficulties, they could communicate via audio with the Students’ site. In most cases, the assistants in Sendai asked local supporters in Kiyosedai to take care of the students. The second type of remote assistance was a Skype based remote assistant. In addition to monitoring the gesture visualization system, the assistants and the students could communicate directly with each other by watching their face in the monitor. For comparison to the present study, we observed the video recorded during the class using telepresence robot, and two previous systems as baselines. We counted the interactions between students and remote assistants, local supporters, and the lecturer while watching the video for 15 min from the beginning of the class.

Table 1 shows the result of the comparison. In the gesture observing only condition, students mainly interacted with local supporters and the lecturer. In the Skype condition, there was a large drop in interaction between students and local supporters and communication between remote assistants increased. Then, for the case of using a telepresence robot, the counted number of interactions between students and the remote assistants increased further. Although we had a small number of participants for this exploration, telepresence robots had the effect of facilitating communication between the remote sites, to some extent.

Table 1. Interaction count with students in different style of remote education. Classified by whom addressed.
Full size table

One remarkable difference between remote education using telepresence robots and the former system is that, during the class, the students became familiar with the robots and remote assistants and made some idle conversation. Figure 12 shows the students taking a commemorative picture with an assistant operating Double and with a lecturer. Additionally, during the exercise period, when students were practicing with their tablet terminal without any trouble, an assistant who was operating Double went for a walk inside the classroom and watched students in other groups or had idle conversation with the local supporters.

Fig. 12.
figure 12

Students enjoying embodied communication with Cogi (left) and Double (right).

Full size image

5 Discussion

From the results, the elderly in remote IT education classes accepted the application of telepresence robots. Telepresence robots also served to better facilitate communication between the students’ side and the lecturer’s side compared with conventional video conferencing styles of remote education. From observations, successful facilitation of communication was found to be subject to the assistant’s outgoing characteristics, not to their IT skills. We adopted Double and Kubi as remote assistant telepresence robots in this study. We could not find a remarkable difference between Double and Kubi, in terms of their practical effect on communication, except that some assistants mentioned Kubi’s tilt motion was useful for observing students’ hands while the students operated the tablet terminal. One particular difference we found was that Cogi and Double had an eye-catching motion so that they could participate in informal communication with the students in spare time, however this was not the case with Kubi. Although evaluations of the impressions of the class were dependent on the condition of the audio connection quality, its affect to the students is getting smaller. We had frequent audio connection trouble especially in the first and the second lessons; students’ evaluation to the class is improving. From the discussion with lecturer side, it was mentioned that it would be useful if assistants could have UI support to distinguish the speaker’s location at the local site.

6 Conclusion

In this paper, we introduced a system using telepresence robots in remote IT education classes for the elderly, in order to facilitate communication between senior communities on the lecturer’s side and the students’ side. From use in an actual class, we confirmed that the telepresence robot is applicable in the class and it also promotes interactions between students and assistants.