Essay on A Survey of Socially Interactive Robots Authors: Terrence Fong, Illah Nourbakhsh, Kerstin Dautenhahn Summarized by: Mehwish Alam

Transcription

1 1 Introduction Essay on A Survey of Socially Interactive Robots Authors: Terrence Fong, Illah Nourbakhsh, Kerstin Dautenhahn Summarized by: Mehwish Alam 1.1 Social Robots: Definition: Social robots are the embodied agents that recognize each other and the environment, they interact with the environment, they learn and interpret the world from their experience (possess history), they communicate and learn from each other. There are two categories of the social robots: 1) Group Oriented Social Robots Group Oriented Social Robots have the behaviour inspiration from insect societies e.g., ant like robots as they search for the food together. They work together so that they can get maximum benefit through group work. In these societies individuals do not matter. Examples: Ant- like robots, distributed robotic systems, soccer playing etc. 2) Individual Social Robots: Such kind of robots are inspired by the creatures which make individual societies e.g., human beings, birds etc. They make relationships with other individuals and keep the social norms and values under consideration. The important factors through which the group oriented social robots differ from the individual robot are social learning and imitation, gestures, cues, natural language communication, emotions, recognition (speech, face), lip reading etc. 1.2 Social Robots and Social Embeddedness: He classified the robots into four classes based on two points How well do they support the social model which is around them How do they deal with the complex interaction scenarios around them? A Survey of Socially Interactive Robots, Cognitive Robotics 1

2 Socially Evocative Robots: These robots have the feelings like human beings. Anthropomorphic robots lie in the category of socially evocative robots. Social Interface: Robots need an interface through which they can easily interact with human and environment and vice versa. As the softwares need a Graphical User Interface so that human beings who are interacting with them can find the communication easier. So robots must use human like gestures and cues in order to provide the natural interface. Socially Receptive: Such robots are socially passive which means that they do not go and interact with individuals on their own rather they copy or imitate the behaviour of the human beings and through it. Sociable: Sociable robots are socially active as they proactively participate in the social interaction. The writers of the research paper have added three following classes in it: Socially Situated: Social robots are meant to interact with the environment and react to it. Socially situated robots must be able to distinguish between other social agents (including human beings) and the objects around them. Socially Embedded: These robots are structurally coupled with the environment and are not fully aware of the human interactional structure. Socially Intelligent: Such robots have the properties of human like social intelligence and social competencies. 1.3 Socially Interactive Robots: By socially interactive robots we do not mean that the communication between robot and human beings is at a distance, it means the face to face Human Robot Interaction (HRI). Social robots must have the following human characteristics: They must be able to use natural cues develop and maintain the relationships build their own personality and behaviour. Express and understand the emotions Communicate with the help of speech They must have the ability to adapt them according to the environment and must have a flexible architecture. 1.4 Why Socially Interactive Robots? We can be used for various therapies for the people of different ages: For the elderly people in hospitals, the robots are trained for their nursing schedules. Robots are also kept there as pet companions to reduce depression and recreate them. Robots can be used as a mobile prosthetic device for the people who are physically retarded. Such robots can be used as tutors, coaches and companions for the children especially during the interaction with the children belonging to the different social groups (e.g., special children) [1] A Survey of Socially Interactive Robots, Cognitive Robotics 2

3 2 Methodology 2.1 Design Approaches There are two design approaches associated with robots, (a) biologically inspired build the robots that copy the behaviour of the living creatures, (b) functionally inspired build such a robot which has desires, feelings and emotions, we should not necessarily know how the mind works. 2.2 Design Issues The traditional issues in designing the robots include cognition (planning, decision making), perception (navigation, environment sensing), action (mobility, manipulation), human robot interaction (user interface, input devices, feedback display) and architecture. Some of the other issues are to build robots that perceive human beings right, they understand the human cues and perform really like human beings. 2.3 Embodiment The more the robot can disturb (take part into) the environment the more it would be embodied. For example, AIBO has 20 actuators in it but KHEPERA in comparison has 2 actuators in it. So, AIBO is more embodied than KHEPERA. For embodiment, the study of the behaviour of living creatures is done (e.g., animals play, fight, run from danger). The robots design can be like human beings, animals, and they must be designed according to their functionality e.g., robots working in hotels should have a tray in place of their hands. During the design of the robots the uncanny valley must be kept into mind. People are heading towards the robot design which is more human like. But if the robot fails to achieve its target of being human like then the disliking increases for that robot. This is what we call uncanny valley. 2.4 Emotion We divide emotions in two categories, discrete emotions and continuous emotions. Discrete emotions are happiness, anger, sadness while continuous emotions are the emotions like reaction to something or excitement. There are artificial emotions attached to the robots which help in providing feedback and the intentions of the robot. Emotions can be expressed through speech, facial expressions and body language. A Survey of Socially Interactive Robots, Cognitive Robotics 3

4 Here I would discuss the comparison between human and robot s emotion model (OCC model): Figure: Overall structure of emotional interaction diagram driven by OCC model and sentiment relation [2] A Survey of Socially Interactive Robots, Cognitive Robotics 4

5 OCC is OCC model is based on interaction related emotion words, which are intuitively assigned from observation of human interaction. event: service task for the inhabitant agent: service agent object: user [3] 2.5 Dialogue Dialogue is an effective way of sharing information. There are many types of dialogues through which robots can interact. First is low level, which includes only the basic names of the objects, human behaviours and names of the body parts. Second, non verbal, which includes gestures, cues. Last, natural language, which means that the robots use the language as the humans do. 2.6 Personality Personality means the behavioural characteristics that distinguish one individual from the other e.g., extroversion, agreeableness etc. extroversion (sociable, outgoing, confidence) agreeableness (friendliness, nice, pleasant) conscientiousness (helpful, hard-working) neuroticism (emotional stability, adjusted) openness (intelligent, imaginative, flexibility) There are five types of personalities that can be attributed to the social robots, tool like, pet, cartoon, artificial being and human like. 2.7 Human Oriented Perception In order to make socially interactive robot the robots must be able to perceive that what human beings are doing, saying etc. They must be able to track the people they are interacting with (people tracking), recognize what people are saying (speech recognition), and recognize the gestures of human beings and also the facial expression of human beings in order to interact effectively with human beings. Facial expression further includes facial detection and recognition and gaze tracking. 2.8 User Modelling A user model generally describes contains attribute that describes a user. Or group of users. Information about the users can be acquired through questions and answers and can be observed. There are many types of user models, cognitive, attentional etc.. User modelling can be qualitative or quantitative. Quantitative user modelling means that the users are divided into subgroups, based on pre defined features. Qualitative user modelling means that Interactional structure analysis, story and scripts are based on subjective aspects of behaviour. 2.9 Socially Situated Learning This is the type of learning in which the robots learn from the environment with which they interact, the basic human competencies and intelligence. It includes social robot learning, which involves transferring of skills, tasks and information, and imitation, which includes copying the behaviours of humans. Socially situated learning includes imitation but there are several questions that are kept in mind for imitation and are as follows: How does the robot know when to imitate? Imitation will be useful when the robot knows when to start or stop imitation and also about the appropriate time. How does the robot know what to imitate? A Survey of Socially Interactive Robots, Cognitive Robotics 5

6 With the help of stream of sensory data robot must decide whether the imitation is relevant to the task or not. How does the robot map observed action into behaviour? Once the robot has learnt the features of the action model, it must be able to produce the same action according to robot s own behaviour. How does the robot evaluate its behaviour, correct errors, and recognize when it has achieved its goal? For the efficiency to be improved robot must be able to recognize when the imitation is wrong and where error has occurred Intentionality Humans use three ways to understand and predict behaviour: physical stance, design stance and intentionality. Physical stance means predictions through physical characteristics, design stance means predictions through design and functionality and intentional stance means systems actions that result from its beliefs and desires. 3 Conclusion Robots increasingly have the potential to interact with people in daily life. It is believed that, based on this ability, they will play an essential role in human society in the not-so-distant future. There would be and still are playing significant role in medical. Scientists have moved a step closer to creating robots that can conduct routine medical procedures on their own. In 2006, there were an estimated 3,540,000 service robots in use, and an estimated 950,000 industrial robots. A different estimate counted more than one million robots in operation worldwide in the first half of 2008, with roughly half in Asia, 32% in Europe, 16% in North America, 1% in Australasia and 1% in Africa. [4] They can also be used in other therapies. They would be able to work as assistants, pets, serving, tourist guides and household staff. Even it can be possible to send the robots to the space for achieving necessary tasks. Social relationship plays a significant role for a robot to determine its appropriate goal position and coordinate its action policy for itself in the multi-user/multi-goal environment. Scientists must not aim at only building the robots that are good at accomplishing specific tasks but social robots must be able to participate in human society to their fullest extent. They must also care for the agreeableness and will of the human beings around them. For this the theories proposed in the given research would be of great help. It would give the insight into for the developers as to what areas must be kept into mind in order to build a robot who must do its specific task as well as engage in the social meetings. A Survey of Socially Interactive Robots, Cognitive Robotics 6

7 References [1] David Feil-Seifer and Maja J Matari c, Defining Socially Assistive Robotics, Interaction Laboratory University of Southern California 941 West 37th Place, Los Angeles, CA [2] Young-Min Kim, Jong-Chan Park, Dong-Soo Kwon, Behavior Coordination of SociallyInteractive Robot using Sentiment Relation Model, Dep. Of Mechanical Engineering KAIST, Guseongdong, Yuseong-gu, Daejeon, Korea [3] Youngmin Kim, Hyong-Euk Lee, Kwang-Hyun Park, and Z. Zenn Bien, Steward Robot: Emotional Agent for Subtle Human-Robot Interaction, The 15th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN06), Hatfield, UK, September 6-8, 2006 [4] A Survey of Socially Interactive Robots, Cognitive Robotics 7