Improvement of Immersion into a Virtual Game World : Using “ Real Robot ” as a Player Character

Immersion is an important sense for a virtual game world. Moreover, Virtual Reality (VR) is an effective method to improve immersion. However, VR requires wearing some devices such as a goggle which can be cumbersome of users. Therefore, an alternative method to improve immersion other than VR is proposed. In this study, the authors propose using a real robot as the player’s character. At this stage, the authors developed a game for experiment called “Catch Apple” in which players can use a real robot. The authors then surveyed 39 research subjects regarding the effect using a real robot had on them. The data was processed using Factor Analysis in order to find factors relating to immersion. The authors compared their factor scores. The results show that the factor scores of the “Real Robot” were significantly greater than the factor scores of the “Virtual Robot” (p < 0.05). In the future, the authors will develop new games and conduct similar experiments to confirm the correctness of the factors relating to immersion.


Introduction
Immersion is one of the most important sensations a player can experience in a game.VR is one gaming design approach to increase the sensation of Immersion.In addition, VR and Immersion are used for mental therapy (1) .Thus, Immersion and VR have been studied by many researchers (2) (3) .However, VR games usually require players to use cumbersome equipment such as oversized goggles.Also, There have been a number of speculative reports of many different types of side effect (4) .For example, many players report experiencing motion sickness when playing VR games.
The authors propose a new method to increase Immersion without relying on VR.The method involves getting the game character out of the virtual world and into the real world.The authors conjecture that a border between the real world and the virtual world becomes ambiguous by using this method.In this study, the authors developed an air hockey game in which the opponent character is a real-world robot.Then, the authors evaluated their Immersion approach surveying research subjects who played this game.

Experimental Equipment
The authors used an air hockey game developed using Unity.Its rules are the same as that of commonplace air hockey.Players operate a mallet to hit a puck.Players earn a point each time the puck is deposited in the goal at the opponent's end.The first player to score four points wins the game.Results are displayed after playing the game.This game has two modes: 1) "Virtual Robot Mode"; 2) and "Real Robot Mode".Players select one of these modes in the title scene at the beginning to start the game.Fig. 1 shows the title screen, the air hockey field and the result screen.

"Virtual Robot Mode"
In "Virtual Robot mode", the opponent character plays within the virtual world (within the display) as in the past.The character is a fashioned similarly after the real robot in "Real Robot Mode".The character operates a mallet by using its own arms in the display.Fig. 2 shows the "Virtual Robot Mode" character in play.
The game calculates the path of the puck by adding the movement distance per frame to the current mallet position.The game moves the opponent's mallet and charact er close to the puck.
The character can say four different types of messages (through the computer speaker) depending on the game situation: 1) "Makenaizo!" (Never give up!) when the opponent gets a point; 2) "Yattazo!"(I did it!)when the character gets a point; 3) "Makechatta …" (I lost…) when the character loses; and 4) "Kattazo ..." (I won!) when the character wins.

"Real Robot Mode"
In "Real Robot Mode", the opponent character appears as a real robot in front of the display.There is no difference between "Virtual Robot mode" and "Real Robot Mode" other than real representation of the opponent character.Fig. 3 shows the "Real Robot Mode" robot in play.This robot was made using Raspberry Pi3, actuators, a speaker, and plastic arms.The robot plays the game by actually using its arms via AirBar.AirBar is a tool that changes a normal display to a touch display.The game calculates a path of the puck by using the same method that is used in "Virtual Robot Mode".The game sends data of the distance between the touch position and the mallet path to Fig. 1 The title screen, the air hockey field and the result screen in the game Fig. 2 The "Virtual Robot Mode" character in play the Raspberry Pi3 via Wi-Fi.The robot moves to approach the puck path.The robot speaks in the same voice as that used in "Real Robot Mode" through a speaker implemented in the robot.The situations when the robot speaks are the same as those when the virtual characters speak in "Virtual Robot Mode".

Experimental Method
The authors compared game users' impressions relating to Immersion that players got from playing in "Virtual Robot Mode" and in "Real Robot Mode".
The authors prepared a questionnaire using the Japanese English Japanses English Table 1 Pairs of polar adjectives and their relation to Immersion Fig. 3 The "Real Robot Mode" robot in play Semantic Differential method.Kanda also used SD method to evaluate an interaction between humans and an autonomous robot (5) .The questionnaire has 26 evaluation items that each consist of a pair of polar adjectives.
Respondents evaluated game impressions according to these evaluation items in 7 levels.The authors determined adjectives related to immersion with reference to Emily Brown's and Paul Cairns' study (6) .Immersion is divided into 3 levels.The first level is Engagement.The second level is Engrossment.The third level is Total Immersion.The authors have considered the possibility that the "novelty" of the game affects the results and, thus, added the polar pairing of "novel-common" to list of adjective evaluation items.Table 1 presents these adjectives and relation to Immersion.The subjects were all Japanese speakers so the survey was conducted in Japanese.However, the authors have translated the adjectives from Japanese into English for the purpose of this paper.Therefore, the adjectives may have different nuances between Japanese and English.Subjects of this experiment were 39 students from the National Institute of Technology, Kurume College (NITKC).
Subjects played the game and responded to the questionnaire after playing the game in respective modes.

Result and Evaluation
The authors assayed the results of the experiment using Wilcoxon signed rank test.The analysis revealed that the results of "Real Robot Mode" were significantly higher than that of "Virtual Robot Mode" (p<0.05).Fig. 4 presents the average of results for each adjective.The authors predicted that these adjectives would have stepwise differences among the Immersion levels.However, the analysis showed that this was not the case.The authors conjecture that the reason for a lack of stepwise differences may be because of a mis-selection of the polar adjectives.Therefore, the authors analyzed the results using Factor Analysis and classified the factors into the immersion levels.
Using Factor Analysis to better understand the results of the experiment, the authors removed adjectives with communalities less than 0.3.This criteria removed four polar adjective pairs: "early-late", "quick-slow", and "gorgeousplain".The authors determined a number of factors to 8 by using eigenvalues and assayed that the number is adequate using the Chi-squared test.The cumulative contribution ratio of these factors is 70.4%.Table 2 presents loadings and proportions of variance and communalities after varimax rotation.The authors regarded loadings greater than 0.5 as high loadings.The high loadings are shown in bold in Table 2.
The authors calculated the factor scores and compared them between "Virtual Robot Mode" and "Real Robot Mode".The authors assayed the averages of the factor scores using Wilcoxon signed rank test.The authors found that the factor scores of "Real Robot Mode" are significantly higher than those of "Virtual Robot Mode" (p<0.05).Fig. 5 presents a comparison of the factor scores between "Virtual Robot Mode" and "Real Robot Mode".This means the "Real Robot Mode" has the effect of giving the player a better impression than the "Virtual Robot Mode".
The authors considered respective factors and classified these into the immersion levels.Factor1 has high loadings in the adjectives of "Preference", "Empathy" and "Atmosphere".The reason for this may be because of a misselection of adjectives in Total Immersion.For example, "casual-serious" are similar to the adjective "Preference".Therefore, Factor1 may relate to "Preference" in Fig. 4 The average of results for each adjective Engagement.Factor2 has high loadings in the adjectives "Preference" and "Rewards".Therefore, Factor2 may relate to "Preference" and "Rewards" in Engagement.Factor3 has high loadings in the adjective "Visual".Therefore, Factor3 may relate to "Visual" in Engrossment.Factor4 has high loadings in the adjective "Want to continue".Therefore, Factor4 relates to "Want to continue" in Engrossment.Factor5 has high loadings in the adjective "Easy of use" in engagement.Therefore, Factor5 may relate to "Easy of use" in Engagement.Factor6 has a high loading in only "wisefoolish".The authors consider that a number of high loadings is not enough to conjecture a relationship between Immersion and Factor6.Therefore, Factor6 may not relate to Immersion.Factor7 has a high loading in only "novelcommon".Therefore, Factor7 may relate to "Novelty".This adjective is added by the authors to study an influence of "Novelty".An influence of Factor7 is low except for "Novelty".Therefore, the authors conjecture that "Novelty" doesn't affect the result of the experiment.Factor8 has no high loadings.However, Factor8 has relatively high positive Fig. 5 Comparison factor scores loadings in "warm-cold" and has relatively high negative loadings in "pretty-detestable".Factor8 may relate to the opponent character since about 60% of the subjects lost when playing the game.When players are immersed in a game and lose, they feel frustrated against their opponent.The frustration doesn't detract from the warmth of the game.The authors conjecture that the player's impression of the opponent relates to the feeling of "Empathy".Therefore, Factor8 may relate to "Empathy" in Total Immersion.
The authors assayed the averages of the factor scores related to Immersion by using the Wilcoxon signed rank test.The authors found that the factor scores of "Real Robot Mode" are significantly higher than those of "Virtual Robot Mode" (p<0.05).Therefore, the authors conjecture that players get more of a feeling of Immersion in "Real Robot Mode" more than in "Virtual Robot Mode".

Conclusions
In this study, the authors attempt to improve the sensation of Immersion in games by placing the virtual game opponent character in the real world as a robot.
Firstly, the authors developed an air hockey game which has two modes: "Virtual Robot Mode" and "Real Robot Mode".In "Virtual Robot Mode", the opponent character exists within the screen; Whereas in "Real Robot Mode", the opponent character exist "in front" of the display in the real world as a robot.
Secondly, the authors experimented with 39 adjectives using the SD method.Subjects played the game and responded to a questionnaire after playing the game in respective modes.The questionnaire has 26 evaluation items that consist of 2 polar adjectives related to Immersion.Table 1 presents these adjectives.
Thirdly, the authors assayed the results of the experiment using the Wilcoxon signed rank test.The authors found that the results of the "Real Robot Mode" are significantly higher than the results of "Virtual Robot Mode" (p<0.05).
Fourth, the authors used factor analysis to better understand the results of the experiment.As a result, 8 factors were found.
Fifthly, the authors assayed the averages of the factor scores using Wilcoxon signed rank test.The authors found that the factor scores of "Real Robot Mode" are significantly higher than those of "Virtual Robot Mode" (p<0.05).
Sixthly, the authors classified respective factors into the immersion level.Table 3 presents the classification of the factors.
Seventhly, the authors assayed the averages of the factor scores related to Immersion using the Wilcoxon signed rank test.The authors found that the factor scores of "Real Robot Mode" are significantly higher than those of "Virtual Robot Mode" (p<0.05).
However, there are three problems.Firstly, the authors only found one factor of Total Immersion.The authors conjecture that the reason is a mis-selection of adjectives in Total Immersion.The authors have to change the adjectives to find factors related to Total Immersion.
Secondly, the factor of Total Immersion doesn't have high loadings.Therefore, the authors conjecture that the subjects couldn't immerse in the game until Total Immersion.Players may have to play the game for a long time in order to immerse until Total Immersion.Moreover, players may have to like the game in order to immerse until Total Immersion.The authors have to develop an enjoyable game.Moreover, the subjects have to play for a longer duration.
Thirdly, because the number of subjects was too few, the results couldn't be analyzed using factor analyses.Moreover, the subjects belonged to the same group.Therefore, the result of factor analysis may be not accurate.The authors have to not only increase the number of subjects, but select a more diverse group of subjects.
Therefore, the authors will experiment again after resolving these problems.

Immersion Levels Immpression Factors
Easy of use Factor5 Rewards Factor2 Visual Factor3 Want to continue Factor4 Total Immersion Empathy Factor8 Engagement Engrossment Preference Table 3 The classification of the factors.
(a) The title screen (b) The air hockey field (c) The result screen in when player lose (d) The result screen in when player win