Collabot: A Robotic System That Assists Library Users Through Collaboration Between Robots

A library serves as a repository of knowledge accessible to individuals of all ages, genders, educational backgrounds, social statuses, and economic levels. It stands as a communal space where community members can gather, bridging information disparities among various societal strata. To enhance accessibility to such libraries for a broader spectrum of people, we have introduced the CollaBot system. This system offers tailored services to users through the collaboration of robots. Our investigation encompassed the acceptance of robot types by users, robot characterization, and the prioritization of robot-provided services. Over the course of three stages of user evaluation, it became evident that participants preferred product-type robots over anthropomorphic robots. Furthermore, they expressed a preference for robots that assist other robots, even if these assisting robots exhibit clumsiness, as opposed to robots that exclusively excel in their designated tasks. Lastly, service prioritization varied based on the specific limitations or deficiencies faced by individual users.


INTRODUCTION
A library is a public space that grants access to books, various materials, and media for all individuals.Library buildings typically offer quiet study areas and communal spaces for group study and collaboration.Libraries can also serve as community hubs, providing programs and fostering lifelong learning for people [1].They are open to everyone, regardless of income level or age, and serve as spaces for learning and information sharing [2].
Materials within libraries are systematically organized according to established classification systems, facilitating quick item retrieval and efficient browsing of collections [3], [4].Furthermore, library systems have been standardized since the 19th century, resulting in highly sophisticated and efficient operations today [5].However, certain individuals may face challenges accessing libraries due to factors such as the library's size, the height of book shelfs, or the weight of books.For instance, libraries efficiently use the Dewey Decimal Classification [6] to organize books, but for general users unfamiliar with this system, locating a desired book among a vast collection using only the book's symbol information can be cognitively demanding.Additionally, since the 19th century, when library systems were standardized, it has become common practice to store library collections in spaces separate from reading rooms [5], [7].Consequently, after selecting a book they wish to read, library users must navigate to the location of the book.While this may be a minor inconvenience when carrying a single book, it can become burdensome when moving to the reading room with several books in tow.
Robots equipped with perception, recognition capabilities and mobility have the potential to provide valuable physical assistance to the users.These robots can perform tasks that may surpass human capabilities, simplify challenging activities, and alleviate burdensome chores, addressing limitations individuals might face.Accordingly, many robotic products have been developed to relieve inconveniences in users' daily lives.Mok et al. (2015) developed and introduced a robotic drawer that collaborates with humans to efficiently help users with their work [8]. Lee et al. (2020) developed a robotic cabinet that assists users in the process of organizing and finding items and verified the effectiveness of this product through experiments [9].Additionally, Sirkin et al. (2015) developed an ottoman that automatically approaches the user sitting on the chair, allowing the user to relax more comfortably [10].Agnihotri and Knight (2019) developed ChairBot, which persuades people to play chess through motion gestures [11].We have confirmed that these robotic products effectively assist users through various related works, and have designed a robotic library system composed of robotic products to assist users in overcoming these challenges through the use of robot technology.To achieve this, we conducted an examination of user behaviors in existing libraries, developed a prototype, carried out a three-stage evaluation, and ultimately created a robotic library system.The design aspects we assessed in the development of this robotic library system to aid users included the robot type, character, and the priority of services provided by the robot.We believe that this study contributes to the Human-Robot Interaction (HRI) community in several ways: Accessibility and Inclusivity: We offer services tailored to users with varying backgrounds, including those familiar and unfamiliar with library usage, as well as individuals of different heights and physical strengths.Our goal is to reduce barriers to library access for users who previously encountered difficulties.
Efficiency and Diversity: Robotic library systems collaborate with each other to recognize users and their surroundings, adapting their services to the specific situation.This adaptable approach allows for the creation of a wide range of services using minimal modules.
Robotic Systems for Public Spaces: This project highlights the potential of robotic systems to enhance the usability and functionality of public spaces.Research in this area can explore how robots can be integrated into various public settings to improve user experiences and services.

Observation and analysis
We conducted observations to understand how users utilize the library, identifying the cognitive and physical difficulties they may encounter.The study involved five users aged between 20 and 40, all of whom held at least an undergraduate degree and did not face any challenges related to vision, mobility, or holding books (See Fig 1).
We asked the users to take out four books they want to read.In this particular library, there was no dedicated book search desk.Instead, users relied on a personal application to search for books in advance.Subsequently, they referenced the classification code information for the book they desired in the library.
We began observing user behavior from the moment they looked for a book in front of the bookshelf.We classified the users' actions into three categories: checking book information in the application (searching and checking the book information by using the smart phone), locating the specific book (moving around and standing still to browse books, and approaching and backing away from the bookcase), and taking out a book (reaching for the book, holding the book, and carrying the book).Users spent the most time locating the specific book.Despite navigating the library sections smoothly, such as 690 -Construction of Building and 720 -Architecture, using the Dewey Decimal Classification, users encountered significant delays in locating specific books within their respective sections.This was observed as they repeatedly moved around to browse books, stood still to browse books, and approached and backed away from the bookcase in the section.Thus, we first developed a bookshelf prototype that shows the location of books.

Prototyping
Initially, we created prototypes of essential library furniture, including a desk, chairs, and a bookcase.Among these, we focused on robotizing the chairs and bookcases.Based on observations conducted during the pre-study phase, we designed the bookcase to display a shelf containing the requested book when users searched for it using the application.Additionally, we designed the chair to move back slightly when a user approached the desk, enhancing seating comfort.

Robotic Bookcase.
To implement the robotic bookcase's motors and control board, we affixed an aluminum profile to the wooden bookcase's rear.For communication with the book search application, we integrated an OpenCR board and a Bluetooth module at the back.During development, we encountered challenges and improved the bookcase's mechanism.Initially, we used a disc-cam system with inclined shelves, but it had limitations.Shelf movement was imprecise due to friction on the slippery board, and informing users of the book's location was challenging.To overcome these issues, we adopted a system with rails, pulleys, and belts.Rails on both sides supported each shelf, with a motor-driven belt and pulley at the base.We tested various pulley, shaft, and belt combinations, using 3D-printed custom clamps to secure them.The embedded microcontroller controls shelf opening and closing (See  .It features a wooden chair with springs at the chair-robot junction mounted on the mobile robot, enabling slight elevation during movement and stable seating when a person sits on it.To maintain LiDAR visibility, there were holes on all four sides, and a 3D-printed component was designed to adjust the LiDAR's height.

STUDY 1: ROBOT TYPE
In the realm of human-robot interaction, the level of anthropomorphism in robot design emerges as a crucial determinant influencing the acceptance of robots [13].Supporters of anthropomorphic robot design contend that robots' human-like features can result in positive user acceptance.This viewpoint is based on the premise that individuals inherently interact with both living beings and non-living entities by using human social cues [14].On the flip side, several studies that suggest that robot designs with lower or non-anthropomorphic features can positively influence user perceptions.Excessive anthropomorphic features, as evidenced by some research [15], may evoke a sense of unease, while robot designs with reduced anthropomorphic elements, as per other studies [16,17], have the potential to alleviate negative responses from consumers.
Users may perceive anthropomorphic robots (AR) offering multiple services positively since users sought diverse assistance from librarians traditionally.However, Weiser (1999) noted that profound technologies blend into everyday life, like product-type robots (PR) integrated into our surroundings.Kwak et al. categorized robotic home services into three types, and participants favored service mediators with non-anthropomorphic PR, indicating a preference for product-like robots over anthropomorphic social robots [18].Therefore, in order to investigate suitable robot types for aiding library users, we developed and compared two distinct types of robots: an AR and PR without being anthropomorphic.
We formulated the following hypotheses: Hypothesis 1. PR (robotic bookshelf and robotic chair), originally designed for storing books and moving objects, are perceived as more useful than the AR.
Hypothesis 2. An AR, characterized by anthropomorphic facial expressions and the ability to assist users as separate and independent entities akin to librarians, are considered more socially engaging than PR.
Hypothesis 3. The evaluation of the overall service provided from the robot vary depending on the type of robot.

Material
This experiment employed Wizard of Oz techniques [19] and utilized two distinct types of robots.
3.1.1Product-type robot.when users selected a book using a dedicated app, the shelf housing the chosen book extended outward, indicating its location and automatically closes in a certain period of time.The robotic bookcase communicates shelf opening frequency information to the robotic chair through ROS communication [12].Based on the information, the robot chair recognizes that the user is holding multiple books and then moves to a designated location where the user can conveniently place the books.Once the user places books on the robotic chair and presses the confirmation button on the application, the robotic chair transports the books to a designated location near the desk.

Anthropomorphic robot.
Temi was used as an AR [20].Temi [20] mapped the experimental space and predetermined locations.Its control program set sequences with movement and facial expressions.The AR displayed book locations on its chest screen when users selected a book.After users placed books on the robot's back shelf, it resembled loading a backpack.The robot then followed the user (See

Participants
The study included 30 participants, consisting of 13 males and 17 females in their twenties and thirties, all with a college education and prior library service experience.

Procedure
Upon their arrival at the laboratory, each participant received a concise explanation of the experiment and completed a consent form, thereby expressing their willingness to participate.Subsequently, participants were instructed to utilize a book search application to locate and select four books, transfer the chosen book to a nearby desk, and then take a seat.Following this, a questionnaire evaluating the robotic experience was administered.Participants encountered both robot conditions in a counterbalanced sequence.

Results
Statistical analysis was applied to all collected data using t-tests.

Findings
The statistical outcomes revealed an overall preference for PR over an AR.In subsequent interviews, most users expressed the product robot's ability to indicate the book's location as more intuitive and practical than the AR's capability (See Table 1).In addition, PR that displayed book locations by using its movement and facilitated book transportation were evaluated more positively than AR that provided a wide array of services, functioning as independent entities akin to librarians.
In contrast to AR, which were engineered to deliver multiple services and assist users across various domains, PR are designed to excel in a singular service domain.Consequently, in order to enable multiple PR to offer users a diverse array of services tailored to specific situations, it becomes imperative to explore the character traits these robots should possess to facilitate user assistance effectively.Therefore, we sought to determine, in Study 2, we investigated the effective robots' characters to assist users.[23,24] are specifically optimized for particular tasks.Unlike multi-function robots, PR find it challenging to offer a wide range of services that cater to diverse user needs.To address this limitation, PR could collaborate with each other.However, when a PR takes on a role beyond its primary function, it may not deliver optimal service.According to the media equation proposed by Reeves and Nass [25], individuals tend to perceive dedicated service providers as more professional and intelligent than those with multiple functions.This experiment aims to determine whether PR should exclusively perform their designated tasks in a professional yet self-centered manner or if they should altruistically assist other products, even with some degree of imperfection.We have formulated the following hypotheses: Hypothesis 1.A PR characterized as altruistic but lacking professionalism will receive higher evaluations for appropriateness in the given situation compared to a PR characterized as selfish but highly professional.This is because the PR characterized as altruistic but lacking professionalism performed a role different from its designated function, which was more suited to the situation.
Hypothesis 2. The PR characterized as selfish but highly professional will be assessed as more intelligent than the altruistic but non-professional PR.This is because the selfish but professional PR is deemed an expert in the tasks it performs.
Hypothesis 3. In terms of social intelligence, the altruistic but non-professional PR will receive higher evaluations than the selfish but professional PR.This is attributed to the altruistic behavior exhibited by the non-professional PR.
Hypothesis 4. The evaluation of services provided may exhibit variations contingent on the character traits exhibited by the robot.

Material
Similar to Study 2, this experiment implemented Wizard of Oz techniques [19] and employed two distinct types of robots.

Selfish but Professional Robot (SPR).
The robotic cart is created by combining Turtlebot [12] at the bottom of a folding cart.We preconfigured two locations for the robotic cart in advance: one near the bookcase and another near the robotic chair and the desk.The location for the robotic chair is set to provide a comfortable position for the user to sit.While observing the user's behavior, a movement command is given remotely via ROS communication to move to the appropriate location based on the situation.
Upon the user's approach to the desk, the robotic chair slightly moves back to provide a comfortable seating position, and when the user departs, it returns to its original position within the desk.Furthermore, the robotic cart receives an order to move alongside the bookcase when they select multiple books.Once the cart is full, it relocates to the desk, enabling the user to transfer the books onto the desk.Subsequently, the robotic cart resumes following the user to assist them when they seek additional books.

Altruistic but Unprofessional Robot(AUR).
We set up two predetermined positions for the robotic cart ahead of time: one near the bookcase and another close to the robotic chair and desk.For robotic chairs, three predetermined positions are designated: one for the user's comfortable seating and two adjacent to predetermined positions for the robot cart.As we monitor the user's actions, we use ROS communication to remotely issue movement commands, ensuring that the robotic cart moves to the relevant location according to the current circumstances.
When the user approaches the desk, the robotic chair moves slightly backward to provide a comfortable seating position.Upon the user's departure, it returns to its designated location within the desk.Additionally, when the robotic cart becomes filled with books, the robotic chair receives a movement command to approach the robotic cart, acting as a supplementary cart to assist the robotic cart.Once the user takes all the books, both the robotic chair and robotic cart are instructed to accompany the user to the desk to assist in transporting the books.

Participants
The study encompassed a total of 30 participants, consisting of 16 males and 14 females in their twenties and thirties.All participants held a college-level education, and had no difficulty reading and writing and using the book search application.

Procedure
Participants were briefed on the experiment's objectives and completed consent forms upon arrival at the lab.They used a book search app to select eight books and move them to a nearby desk.Due to the robotic cart's capacity limitation (four books), participants in the SPR chair scenario transported four books first, then made a second trip for the remaining four books using the robotic cart.
In contrast, with the AUR chair, all eight books were moved together, acting as an auxiliary cart when the robotic cart was full.After book transport, the chair adjusted for comfortable seating and returned to its original position when the user left.A postexperiment questionnaire assessed the robotic experience.Participants experienced both robot conditions in a counterbalanced sequence.

Results
Statistical analysis was conducted on all collected data using t-tests.
Hypothesis 2 was not supported by the data (F = -3.871,p < .001).The perceived intelligence of the AUR chair, which assisted the robotic cart, was rated higher than that of the SPR chair that carried out its duties professionally (MSPR = 5.08, SDSPR = 1.18;MAUR = 5.74, SDAUR = 0.96).

Findings
A robot that altruistically helped the other robot was evaluated more favorably than the robot that did not in terms of appropriateness, intelligence, and social intelligence.Descriptive data also implied that the AUR was evaluated as providing better service than the SPR.As in Study 1, interviews were conducted (See Table 1).In the case of AUR, books sometimes fell to the floor while moving, and sometimes lightly bumped into a bookcase due to difficulties in positioning themselves between the robotic cart, the bookshelf, and the user.Nevertheless, in the interviews, participants responded that although the service quality provided by the AUR was low, they had a positive impression of it helping the other robot and struggling to reduce users' efforts.
As libraries are spaces used by many people, it is likely that a single robot may need to assist multiple users simultaneously.Additionally, due to physical limitations, a single robot cannot provide multiple services simultaneously.Therefore, additional research was conducted to determine the priority of services the robot should provide to the user when the user requires multiple services simultaneously.

STUDY 3: THE PRIORITY OF SERVICES PROVIDED BY THE ROBOT
In Study 2, it was established that a robot assisting other robots, even if lacking in professionalism, was received more positively than a SPR.Practical constraints limit the number of robots available in public spaces.Moreover, in such spaces where diverse individuals congregate, multiple users may simultaneously require robot assistance.In such scenarios, accommodating multiple robots to assist one user can be challenging, and there are inherent physical limitations in one robot simultaneously providing multiple services.Although various robotic furniture designs have been developed to collaborate with the users [8,9], there is a notable gap in research that explores how to create services through the combination and collaboration of robotic furniture within limited spaces and with constrained resources.Therefore, in this study, we investigated the prioritization of services when a user necessitates multiple services simultaneously.
Robotic systems within a library context were categorized into two distinct types: those providing services deemed critical for achieving the user's desired outcome and those offering services that, while not essential, enhance user convenience.
Hypothesis 1. Users' perceptions of the utility of a robotic system will differ based on task success.Users who, due to their own limitations, achieve something they could not have without the robot's assistance, will evaluate the robotic system providing indispensable services more favorably than the one offering nonessential services.Conversely, users who accomplish their objectives regardless of whether the robot provides the service may not discern a difference in service or may rate the robot providing relief from inconvenience as more useful.
Hypothesis 2. Perceived social intelligence of robotic systems will be assessed differently based on task success.Users facing hindrances in achieving their goals, such as users with limitations, may appraise a robotic system that provides services to help them overcome their limitations more positively than a robotic system addressing an inconvenience.In contrast, users without such a handicap may evaluate the same services differently.This differentiation stems from the perception that, in the case of users with limitations, they would consider the robotic system that provides essential services for them as effectively cooperating to perceive the user and the situation, offering appropriate services that recognize and respond adeptly to the circumstances.
Hypothesis 3. The evaluation of robotic systems will vary depending on task success.Users who could not complete a task without the assistance of a robotic system may assess more favorably the robotic system they believe provided essential services.Conversely, users capable of task execution without a robotic system may express a positive evaluation of a robotic system that relieves them from laborious tasks.

Robot System Providing a Service Deemed Critical for
Achieving the User's Desired Outcome.When a user selects a book through the application, the location information of the shelf containing the book is transmitted to the bookcase via Bluetooth communication and then to the control PC through ROS communication.If the control PC determines that the user has selected a book from the top shelf of the bookcase, it commands the robotic chair to move to the bottom of the shelf via ROS communication.Users can then use the robotic chair as a ladder to reach the book.Once the user descends from the chair, the control PC instructs the robotic chair to return to its initial position.Additionally, when the user approaches the desk holding a book, another robotic chair located under the desk slightly moves outward, allowing the user to sit comfortably.

Robot System Providing a Service Not Essential, Enhance User
Convenience.When a user chooses a book using the application, the shelf's location information is sent via Bluetooth communication to the bookcase, resulting in the shelf opening.Simultaneously, this information is also sent to the control PC through ROS communication.The control PC keeps track of the total number of times the shelf has been opened.If the control PC determines that the user has selected multiple books, it directs the robotic chair to move beside the bookcase.Users can then use the repositioned robotic chair as a cart.Once the user places several books onto the robotic chair, the robotic chair transports the books to the side of the desk.Furthermore, as the user, accompanied by the robotic chair serving as a cart, approaches the desk, another robotic chair under the desk adjusts its position slightly backward to offer a more comfortable seating experience and returns to its original position when the user leaves.

Participants
The study included 30 participants, 11 males and 19 females in their twenties and thirties.In the pilot test, two participants, one measuring 160 cm and the other 163 cm, couldn't reach the top shelf, while two participants taller than 165 cm, one 165 cm and the other 166 cm, successfully retrieved books from the top shelf.In the main experiment, there were 30 participants, with 15 below 165 cm and 15 above this height.In addition, all participants were people who had no difficulty finding, taking out, holding books, and moving.

Procedure
After receiving a briefing about the experiment, participants completed a written consent form to indicate their willingness to participate.Subsequently, in all conditions, they were instructed to select a total of four books: two books from the top of the bookshelf and two books from the middle.
In the condition where the robotic chair functioned as a ladder, whenever the user attempted to retrieve a book located at the highest point, the chair approached the book's location, allowing the user to step on it and easily access the book.Conversely, in the condition where the robotic chair acted as a cart, when the user selected four books, it positioned itself next to the bookshelf, enabling the user to load the books onto it for transport to the desk.
Following their experience in each condition, participants were asked to complete a survey evaluating the robotic system they had interacted with.All subjects experienced both conditions in a counterbalanced order.

Results
Three participants shorter than 165 cm resorted to jumping to successfully retrieve the books.Consequently, the participant group was categorized into two subsets: those who managed to retrieve both books (comprising 18 individuals) and those who were unable to retrieve a single or both book (comprising 12 individuals).Statistical analysis was conducted on all collected data using twoway ANOVA.
Hypothesis 1 was supported by the data (F = 8.018, p = .008).Participants who were unable to retrieve a book from the bookshelf without the ladder service provided by the robotic chair found the robotic system offering the ladder service to be more useful than the one providing the cart service (Mladder = 6.36,SDladder = 0.66; Mcart = 5.33, SDcart = 1.24).Conversely, participants who successfully retrieved both books from the top of the shelf, even without the ladder service, perceived the robotic system offering the cart service as more useful than the one providing the ladder service (Mladder = 5.81, SDladder = 0.94; Mcart = 6.02,SDcart = 0.73).
Hypothesis 2 was not confirmed by the data (F = 2.987, p = .098).There was no significant difference in the perceived social intelligence of the two types of robotic systems based on user groups.
Hypothesis 3 was substantiated by the data (F = 7.610, p = .010).Participants in the group that encountered difficulty retrieving a book from the bookshelf when the robotic system did not provide the ladder service tended to evaluate the robotic system offering the ladder service more positively than the one providing the cart service (Mladder = 6.27,SDladder = 0.68; Mcart = 5.56, SDcart = 1.18).Conversely, participants in the group that successfully retrieved the two books from the top, even without the ladder service, rated the robotic system offering the cart service more favorably than the one providing the ladder service (Mladder = 5.74, SDladder = 0.75; Mcart = 5.94, SDcart = 0.88).

Findings
Individuals who encountered difficulties when trying to retrieve books due to their height tended to evaluate the robotic system that assisted them in overcoming these limitations and accomplishing tasks using the ladder service more positively than robotic systems that undertook cumbersome tasks.Conversely, individuals in the group who did not face difficulties in accessing books assessed the robotic system providing the cart service, which undertook the cumbersome task, more favorably than the ladder system.
During interviews (See Table 1), participants in the group who did not encounter difficulties when trying to access books located at the top of the bookshelf mentioned that they often had to step away from the bookshelf to view the book titles.This was because the books were positioned above their eye level, making it challenging to read the titles from a close distance.Despite the task of retrieving the books being somewhat challenging due to their height, these participants were able to successfully complete it without requiring assistance from the robotic system.Consequently, they expressed that, much like the cart service, the ladder service was not considered a necessity for them.
The findings of this study suggest that the evaluation of the same service can vary based on the user's circumstances and context.This underscores the importance of designing a robotic system capable of discerning the user's priorities and delivering services accordingly.Consequently, we have developed CollaBot, a system that tailors its services through collaboration between robotic products, adapting to the user's condition and the specific situations they encounter.

Design Implications
Through a series of experimental studies, we have distilled three key design considerations for the development of CollaBot.
The first design implication underscores the importance of retrofitting existing products with robotic features.Participants expressed a more favorable evaluation of the PR, which represents the integration of existing products with robotic features, compared to an AR operating as an independent entity.Consequently, CollaBot should be designed to facilitate the collaboration of multiple PR in assisting users.
The second principle advocates for occasional role changes among robots to assist their counterparts.Users evaluated robots that provided assistance to other robots, even when such assistance was accompanied by clumsiness, as offering better services, displaying higher intelligence, demonstrating superior social intelligence, and delivering superior service compared to robots that did not engage in such collaborative assistance.Therefore, fostering collaboration among robots and enabling them to provide services through mutual aid can enhance the overall user experience.
The final principle focuses on providing customized services.Participants exhibited distinct preferences for services based on their specific limitations or deficiencies.Therefore, CollaBot's design must carefully address users' most pressing needs and deliver tailored assistance accordingly.

COLLABOT
Following the three design principles, we developed a robotic library, an integrated system featuring a dedicated application and multiple robotic components: a robotic bookcase and robotic chairs.These robotic components collect environmental and user data, collaborate to recognize contextual cues, and deliver tailored services to users.Each robotic component can perform its primary function but also adapt its role as needed (See Fig 4 and 5).

Interaction Flow
We have delineated the interaction flow among the user, the robotic bookcase, and the robotic chair, focusing on user behaviors, as follows:  Retrieving a Book: The robotic bookcase assesses the user's height and the book's location.If it identifies that the user is facing difficulties retrieving the book, it issues a command to the robotic chair, instructing it to assist the user by functioning as a ladder.In cases where the user's height allows for unaided book retrieval, no commands are issued to the robotic chair.
Retrieving Multiple Books: When the robotic bookcase detects that the user has selected multiple books and perceives that the user might encounter challenges in handling them simultaneously, it directs the robotic chair to help, serving as a cart.The robotic chair positions itself beside the user, allowing the user to place books on it conveniently.
Leaving the Library: Once the user has selected all the desired books and proceeds to exit the library, the robotic chair, now laden with books, follows the user to the reading room or lending area.When a user approaches the desk, the chair gently moves backward to facilitate a comfortable seating experience.Subsequently, upon the user's departure, the chair reverts to its initial position (See Fig 5).

System Overview
A robotic library is an integrated system that includes dedicated applications based on Bluetooth technology using MIT App Inventor [32].and several robotic things, such as a robotic bookcase and robotic stools.The robotic bookcase is equipped with rails, motors, belts, a built-in OpenCR board, and a Bluetooth module.The robotic stools have mobile robots built into their bases.The robotic library system integrates information, recognizes situational context, and provides customized services to users through ROS communication and Bluetooth communication.

User Height Recognition.
We experimented with several prototypes to measure users' height.Initially, we attempted to estimate height using a robotic gate equipped with ultrasonic sensors placed at equal intervals.However, this method had limitations in that the accuracy of the ultrasonic sensor was low, and it could only provide a height range rather than an exact value.
Subsequently, we installed an Azure camera on top of the bookcase to extract the user's skeleton for height calculation.Unfortunately, this method proved less effective because real-time height changes resulted from continuous updates to the user's posture.Finally, we successfully recognized the user's height by calculating the distance from the ground to the user's head using the Azure camera.

Bookcase Shelf
Control.We designed a mechanism in which the bookcase automatically closes the shelf when it detects that the user has pulled out a book from the extended shelf.
The Region of Interest (ROI) for each bookcase shelf is defined using the RGB camera located on the top of the bookshelf.The image similarity of the set ROI is calculated using Structural Similarity Index (SSIM) [33].When the calculated similarity exceeds a specified threshold, the bookcase shelf is closed.

Central Control of Multiple Robot.
For seamless communication among robotic components, the central control PC receives and integrates all information, determines the system's context, and manages services.The Bluetooth module, connected to the book search application, transmits information about the shelf location and shelf opening frequency to the control PC through the OpenCR.Information from the cameras can be processed immediately by the control PC because the cameras are directly connected to it.By integrating all the information generated by the CollaBot system, the central control PC publishes a ROS topic related to the movement of the robotic chair.

Localization of Multiple Chairs.
To operate multiple chairs, we assign a namespace to each Turtlebot [12] within every chair.The CollaBot environment space is mapped using Turtlebot 3 SLAM [34], and upon receiving movement commands from the control PC, each chair moves to its respective location on the map.

GENERAL DISCUSSION
In this study, our goal was to create CollaBot, a robotic system designed to improve library accessibility.In this section, we'll explore key insights, future prospects, and limitations from the CollaBot development process.

Summary and Interpretation
Participants exhibited a preference for PR over AR with a separate and distinct presence.Additionally, they more favorably evaluated robots that collaborated with other robots, even when these robots occasionally exhibited clumsiness, as compared to robots that solely performed their designated tasks in isolation.Furthermore, users' preferred services varied based on their specific limitations or deficiencies.
An intriguing result in Study 1 was that PR were assessed as more socially adept than AR.This finding contrasts with previous research where AR were considered more socially engaging when interacting with humans [35], [36], [37].This suggests that unless users engage in social interactions with the robot, such as engaging in casual conversations or seeking advice, anthropomorphism may not significantly impact a robot's perceived sociability.
Conventional Human-Computer Interaction (HCI) studies have posited that viewers tend to place greater intellectual trust in information delivered by specialized channels that focus solely on one topic, rather than channels covering a multitude of topics [38].However, in Study 2, participants evaluated the robot that assisted other robots as more intelligent, despite its occasional errors when performing tasks beyond its primary function, compared to the robot solely focused on its assigned tasks.This discrepancy could potentially be attributed to the nature of the tasks involved.Reporting and delivering news require expertise, and the consequences of mistakes in such tasks can be critical.Conversely, the transportation of books within a library entails lower training requirements and does not pose significant risks to users even if occasional errors occur.This suggests that the acceptability of a robot handling multiple roles may vary depending on the task type.Consequently, it is imperative to conduct future research to assess whether users view role-changing robots positively based on task types and situational demands.

Limitation and Future Work
The operation of a mobile robot carries inherent risks, such as the possibility of collisions with users or potential accidents, especially when users attempt to climb onto the robot unexpectedly.Therefore, we plan to undertake future research with a primary focus on ensuring the safety of the robot's shape, structure and interactions.
Given that libraries typically maintain a tranquil ambiance, we intentionally avoided using auditory cues to inform the user of the book location.Instead, we leveraged the robot's mobility to visually and tactically indicate the location of books by extending the shelf containing the desired book.Nevertheless, there may be more effective interfaces for communicating the book's location.Thus, we plan to conduct research on various interfaces that notify the location of books, such as an interface where an LED lights up on the bookshelf where the book is located, or an interface where the smartphone vibrates when the book is approached.
This study has introduced a robotic library system aimed at enhancing information accessibility for a broader range of people.To enhance the value of this study, we intend to augment it by incorporating additional design guidelines catering to vulnerable groups, including the elderly and children.This will involve designing and conducting experiments that offer more comprehensive and detailed services tailored to the unique needs of these specific user demographics.The study received approval from the Institutional Review Board (IRB) at KIST (KIST-202209-HR-022).

Figure 2 :
Figure 2: Iterative improvement to the functional mechanism of the robotic bookcase, (left and middle) disccam mechanism prototype, (right) 3D-printed custom clamps prototype 2.2.2 Robotic Chair.The robotic chair design is built upon the Turtlebot3 waffle-pi, a ROS-based mobile robot [12].It features a wooden chair with springs at the chair-robot junction mounted on the mobile robot, enabling slight elevation during movement and stable seating when a person sits on it.To maintain LiDAR visibility, there were holes on all four sides, and a 3D-printed component was designed to adjust the LiDAR's height.

Figure 3 :
Figure 3: Temi used as an anthropomorphic robot

Figure 5 :
Figure 5: CollaBot Book Search: When a user selects a book via the application, the robotic bookcase extends the shelf containing the book and notifies the user of its location.Retrieving a Book: The robotic bookcase assesses the user's height and the book's location.If it identifies that the user is facing difficulties retrieving the book, it issues a command to the robotic chair, instructing it to assist the user by functioning as a ladder.In cases where the user's height allows for unaided book retrieval, no commands are issued to the robotic chair.Retrieving Multiple Books: When the robotic bookcase detects that the user has selected multiple books and perceives that the user might encounter challenges in handling them simultaneously, it directs the robotic chair to help, serving as a cart.The robotic chair positions itself beside the user, allowing the user to place books on it conveniently.Leaving the Library: Once the user has selected all the desired books and proceeds to exit the library, the robotic chair, now laden with books, follows the user to the reading room or lending area.When a user approaches the desk, the chair gently moves backward to facilitate a comfortable seating experience.Subsequently, upon the user's departure, the chair reverts to its initial position (SeeFig  5).

Table 1 .
Questions for Semi-Structured Interview Three Main Questions for Semi-structured Interview • Have you ever used a robot?• What do you think is the difference between the two robots?• Which robot helped you more effectively?-Why do you think so?