Psychological strategy board games offer players the enjoyment of inferring opponents’ intentions through subtle cues such as facial expressions and gestures. However, reproducing such experiences in solo play remains challenging. This study investigates whether a small communication robot with limited expressive capabilities can reproduce such psychological interactions. Through a series of experiments, we found that the robot’s emotional reactions affected participants’ in-game choices and significantly altered their perceptions of the robot’s intelligence, difficulty level, and the overall enjoyment of the game. These findings suggest that even with limited expressive capabilities, nonverbal cues from a robot can meaningfully simulate psychological gameplay interactions, offering new possibilities for engaging solo experiences with artificial agents.

Homeostasis is a self-regulatory process, wherein an organism maintains a specific internal physiological state. Homeostatic reinforcement learning (RL) is a framework recently proposed in computational neuroscience to explain animal behavior. Homeostatic RL organizes the behaviors of autonomous embodied agents according to the demands of the internal dynamics of their bodies, coupled with the external environment. Thus, it provides a basis for real-world autonomous agents, such as robots, to continually acquire and learn integrated behaviors for survival. However, prior studies have generally explored problems pertaining to limited size, as the agent must handle observations of such coupled dynamics. To overcome this restriction, we developed an advanced method to realize scaled-up homeostatic RL using deep RL. Furthermore, several rewards for homeostasis have been proposed in the literature. We identified that the reward definition that uses the difference in drive function yields the best results. We created two benchmark environments for homeostasis and performed a behavioral analysis. The analysis showed that the trained agents in each environment changed their behavior based on their internal physiological states. Finally, we extended our method to address vision using deep convolutional neural networks. The analysis of a trained agent revealed that it has visual saliency rooted in the survival environment and internal representations resulting from multimodal input.

This paper presents a new equivalent-input-disturbance (EID) approach that is the combination of the EID approach and model-predictive-control (MPC) method. The EID approach estimates the influence of a disturbance and uses it to reject the disturbance. Thus, the amplitude of the control input can be beyond the limitation if the original disturbance is large. To solve this problem, this paper combines the EID approach and the MPC method. The MPC method adjusts the amplitude of an estimated EID during a large disturbance. A numerical example uses a structural model and an earthquake. The results of it verify the validity of our method and the presented method has a better control performance than the previous EID approach. Furthermore, the MPC method adjusts the control input lest it is beyond the limitation.

The embodied autonomous agent, such as a house-hold robot or a pet robot, is required to satisfy multiple requirements simultaneously. One possible approach would be to train the agent to simultaneously solve multiple tasks using a single physical entity (i.e., the body). However, integrating multiple tasks is not a trivial problem in the reward design paradigm of reinforcement learning (RL). Homeostatic RL treats the integration of multiple tasks as a control problem over the agent's internal bodily state. Still, learning in previous studies was extremely slow. In this study, we report novel behavior switching architectures: the Interoceptive Mixture of Experts (IMoE) and the Interoceptive Behavior Switching (IBS) for homeostatic RL agents with continuous motor control. In these architectures, the agent switches between multiple policies using internal body states (interoception). We tested IMoE and IBS in four homeostatic RL environments. We also compared IMoE and IBS with a fully connected model and a query key value switching model with full observation policies. The results indicate that the proposed architectures provide better or competitive results in all four benchmark environments.

This paper improves the habitability of a building for wind load using an equivalent-input-disturbance (EID) approach. In these few decades, many high-rise buildings are constructed around the world. The influence of wind load is large for high-rise buildings, and it causes a decrease in habitability. Habitability relates to the acceleration of a structural response. To improve the habitability of a building for a wind load, we use a lagged-derivative function and EID approach. An EID approach estimates the effect of a disturbance on an output signal. Thus, an acceleration has to be output to estimate an EID. This paper uses a lagged derivative to output the acceleration. The numerical verification uses a building model of 240m and demonstrates that the presented method suppressed the absolute acceleration and improves the habitability of a building to a wind load.

There are various studies on multimodal internal state expression by robots and virtual agents. Some communication robots, especially soft robots such as stuffed toy robots, are often used for intimate communication with physical contact. In this research, we focused on "internal desire" as one of the internal states of an agent, and discussed a method of expressing desire by combining voluntary behavior and heartbeat expression as an involuntary physiological phenomenon. In this experiment, first, we verified the expression of desire by behavior or heartbeat expression individually. Next, we verified how the strength of internal desire and the suppression of desire change through a combination of agent behavior and heart rate, and we analyze their relationship. The results showed that both desire behaviors, which are gazing behavior and approaching the object, and the different patterns of heart rate changes affected the intensity of desire. In addition, the combination of large heart-rate changes and low desire-level behaviors expressed the agent's suppressed desire state.

Humans observe the behavior of others and exhibit conforming behavior according to the situation. In this study, we propose a method to change the user's walking behavior and attention by changing the ratio of multiple virtual agents in action. We verified whether the attention to a poster on the wall and the walking behavior of a pedestrian changed according to the type and proportion of actions exhibited by multiple agents presented to a particular pedestrian. The results showed that the participant's walking speed and attention to the wall advertisement changed according to the agents' walking behavior, such as deceleration and stopping.

In this paper, we propose a system to enhance the user's learning motivation by showing multiple anthropomorphic agents as members of outgroups that are competitors and an agent that cooperates with the user in learning and working as a member of the internal group. To stimulate competition by changing the rank fluctuation of the group, the agents show their activeness in the learning contents and group discussions. Four kinds of action animations of "work,'' "break,'' "look at other groups,'' and "speak within the group'' were prepared, and each agent performs the actions for each duration corresponding to the progress of the group. The results suggested that the rank fluctuation elevated competitive spirit and learning motivation.

This research aims at the emotional expression of a robotic hand through various gripping manners on the user's hand. The proposed system is implemented with a robotic hand's various haptic actuators to realize the change of the fingers' gripping force and the robotic hand's holding duration so that the user can haptically estimate the emotion of the robot. The system is expected to provide stress relief or emotional stability, especially for elderly or challenged people. Our experiments showed that the stronger gripping force was perceived as more hypersensitive and higher affinity and that a longer grasping time increased the higher affinity.

In this paper, we verify whether an agent's internal state of desire can be expressed by combining the agent's visible behavior and tactile heartbeat as an involuntary physiological phenomenon. The results of the experiment confirmed that the agent's desires related to food are expressed by both heartbeat and behavior, furthermore, their combination expresses a strong desire. On the other hand, "being interested," the initial state of desire, and "being patient," internal suppression, both appear in the heartbeat but not in visual behaviors.

In this paper, we introduce our VR system which uses projection, speakers, and a fan to express the distinctive experience of concentration in a Japanese tea ceremony. This system aims to broaden the participant's field of perception (view, hearing, and wind) from a narrow field of view in the tea bowl to multiple sensations of the outside of the tea ceremony room. The continuous concentration, which is presumed by the user's stable motion while mixing tea using a bamboo tea whisk, broadens the field of view and other senses. The accelerometer sensor values of a bamboo whisk were used in an auto-correlation function for detecting the user's stable mixing motion in the tea bowl. In the demonstration experiment in IVRC 2013 (International Virtual Reality Contest 2013), the participants commented positively on the system.

In this research, we aimed to encourage a common recognition of object ownership on the part of agents and humans to elicit more human empathy for the agent. To do so, we propose an ownership-desire model in which the agent's behaviors are generated based on its internal states of attention, interest and desire for ownership. We conducted an experiment to verify the effectiveness of the designed behaviors corresponding to our proposed model, and the results showed that all behaviors corresponding to the internal states were perceived as we designed. We conjectured that the behavioral durations of the agent's attention and interest would affect the guesses of the agent's internal states.

We propose an attention-attracting method for wall advertisements by using a virtual agent whose walk precedes passersby from the viewpoint of conforming behaviors. Conforming behaviors such as adapting opinions and actions to surrounding people are often seen in human communications. We focused on these psychological responses and designed a system that projects a virtual agent's walking animation on the wall surface around an advertisement in order to cause users to become interested in the advertisement. In this study, we verified whether a virtual agent provokes the user's conformity in walking behavior, and discussed whether the agent's walking behaviors indirectly attract the user's attention to a wall advertisement, changing his or her impression of the advertisement. The results suggest that the agent's walking speed and various behaviors toward the advertisement affects the passerby's walking behavior, his or her attention and interest on the advertisement, and the impressions of the advertisement itself.

In our past study, we have aimed at emotional expressions of a robot using artificial physiological phenomena, such as breathing and heartbeat, and we demonstrated that both the robot's respiratory rate and heart rate could express the arousal. On the other hand, in previous works, it was not clear whether these physiological phenomena affected the valence (positive -- negative) level of the robot's emotion. In this paper, we also examined other physiological aspects of breathing and heartbeat, specifically the inspiration / expiration ratio and heartbeat variability. Using a robot that was designed based on actual animal respiration and heartbeat mechanisms, we verified whether these physiological changes affect emotional valence in the robot's internal state. The results showed that the robot's emotional valence changed with the interaction of the respiration rate and inspiration / expiration ratio, suggesting the possibility of measuring the robot's emotional expression with two axes, arousal and valence.

In this research, we propose a system that shows a group of multiple agents as other learners to the e-leaning user. For continuous learning in e-learning, promoting the user's motivation for learning and appropriate breaks are considered to improve the user's work efficiency. The proposed system controls the proportion of the agents' behaviors in their group by allocating the state of study--either studying or taking a break--for each agent in order to indirectly indicate concentration or appropriate breaks. Consequently, the system offers multiple actions, but the user has room to make the choice regarding the action. Depending on the concentration degree and the duration of study of the user, the necessity of a break is calculated. That is used to control the ratio of agents who have a break and to avoid the forced feeling resulting from only one action being presented by a single agent. In this paper, we examined how the ratio of behavior in the agent group influences the user. As a result, a greater ratio of studying to resting in the group influenced the participants' behaviors.

One of the important factors in medical and nursing care recently has been arousing the emotions of patients. Many types of communication robots and pet robots have been developed as communication partners for patients. The user's emotions are stimulated during direct communication with the robot, although there is less chance that the robot will approach the user in other situations such as watching TV and listening to music without disturbing him or her.The user feels that the robot is troublesome during the user's other tasks. The purpose of this research is to elevate the user's emotional experience through emotional expression by physiological expressions of a partner robot in the user's daily life. Ambient but emotional expressions of physiological phenomena are perceived by touch, even when the user is concentrating on other tasks. First, we focused on breathing, heartbeat and body temperature as the physiological phenomena. From the results of the evaluations of the robot's heartbeat and body temperature, along with our previous results for the breathing, each expression has arousal and pleasure axes of the robot's situation. In this paper, we focus on The joint attention of the robot and user to an emotional photograph, and we verified whether the strength of the user's own emotional response to the content was changed by the physiological expressions of the robot while they looked at photographs together. The results suggest that the physiological expression of the robot would make the common emotional experience of users the user's own emotion in the experience more excited and more relaxed.

This paper proposes an interactive manager agent system that adjusts schedule among multiple persons through negotiations on behalf of a human organizer, such as a steward or team manager. In order to increase the users' cooperative decision-making, we implemented the manager agent that negotiates with each member on the schedule using 1) the suitable attitude depending on current situation and partner of negotiation and 2) the graphical representation of background negotiations between the agent and the other users: a) the other users' icons, b) the blinking balloons of both the other users and the agent, and c) the agent's looking-back actions for the other users' icons. The results of the evaluations showed that both the individual use and the combination of the proposed expressions could make the participants feel i) as though there were under-going negotiations between the agent and the other users and ii) as though the agent had both the effort and compassion to the user. Moreover, the user's cooperative decision-making was increased by the proposed method.

In this research, we focus on physiological phenomena expression of robots, propose the Body-Emotion Model (BEM), which concerns the relationship between the internal states of robots and their involuntary physical reactions. We expect that the robot will be able to express the delicate nuances of its intention by controlling robots' emotions and feelings with each individual parameter based on BEM. We propose a stuffed-toy robot system, BREAR, which has a mechanical structure to express breathing, heartbeating, temperature, and bodily movement. The breathing mechanism commonly controls the abdominal motion, breathing motion, and air flows of breath. In this paper, we focus on robot's breathing, verify two subjective evaluations: 1) the user's sense that the robot was alive and 2) the perceived states of the robot based on its breathing speed. The results showed that our proposed method of breathing expression can show a state of living and that the breathing speed was interpreted as the robot's emotion of arousal.

In this paper, we propose spatial communication between a virtual agent and a user through common space in both virtual world and real space. For this purpose, we propose the virtual agent system SCoViA, which renders a synchronized synthesis of the agent's appearance corresponding to the user's relative position to the monitor based on synchronization with the user's motion parallax in order to realize human-agent communication in the real world. In this system, a real-time three-dimensional computer-generated (3DCG) agent is drawn from the changing viewpoint of the user in a virtual space corresponding to the position of the user's head as detected by face tracking. We conducted two verifications and discussed the spatial communication between a virtual agent and a user. First, we verified the effect of a synchronized redrawing of the virtual agent for the accurate recognition of a particular object in the real world. Next, we verified the approachability of the agent by reacting to the user's eye contact from a diagonal degree to the virtual agent. The results of the evaluations showed that the virtual agent's eye contact affected approachability regardless of the user's viewpoint and that our proposed system using motion parallax could significantly improve the accuracy of the agent's gazing position with each real object. Finally, we discuss the possibility of the real-world human-agent interaction using positional relationship among the agent, real objects, and the user.

This paper proposes an effective communication with an agent's appearance and reality of the acceptance level of the user's order using verbal and nonverbal expressions. The crossmodal combination is expected to enable delicate expressions of the agent's internal state especially when the agent's decision of the acceptance is different from the agent's mind, so the user becomes aware of the difficult situations. We have proposed to adopt the facial expression and the flexion of the word ending as the parameters of the agent's internal state. The expressions are attached to the scripts of each acceptance level. The results of the expressions showed the following: 1) the effectiveness of both the facial and flexion expressions, and 2) the crossmodal combinations that generate the agent's concealed but real feeling.

In this paper, we propose a virtual agent system that introduces the current news on the Webs depending on the user's reactions to the understanding. We focused on the gesture of the user to estimate the user's state of understanding based on the head motion. The goal of the research is the user's under- standing of the news regardless of the knowledge level. The system controls 1) the level of the news content corresponding to the user's gesture of inclining her/his head and 2) the audibility of the agent's speech (the volume and the speed) by the user's head motion to turn her/his ear to the agent. The results of the evaluations showed the effectiveness of the interactive change of the reading behaviors to be understandable and audible.

In this paper, we propose an involuntary expression of embodied robots by adopting goose bumps. The goose bumps are caused by not only external stimuli such as cold temperature but also the internal state of the robot such as fear. For more natural anthropomorphism, the combination of involuntary and voluntary expressions should enable realistic animacy and life-like agency. The bumps on the robot's skin are generated by changing lengths of thin rods from each hole. The lengths are controlled by a servo motor which pulls nylon strings connected to the base of thin rods.