2.1.1.1. Humanoid head

The most effective anthropomorphic feature of a robot is its head. To project human likeness and better expressiveness, the simplest kind of humanoid robot heads are equipped with RGB LEDs, cameras, microphones and speakers. These mechanical parts are mostly cost-effective and provide a variety of expressions for more naturalistic human robot interaction. DARwIn-OP, HOAP-3, Pepper, NAO, UXA-90, Roboy and ASIMO are some of the examples of humanoids with faces equipped with LEDs and speakers. Perception of emotions by humans, while interacting with these robots is at times difficult due to limited modes of expressions giving unrealistic or mechanical effect. Some humanoids are provided with kinematic heads. These humanoid heads can perform transformations from one emotional state to another by tilting head, moving eyes and mouth etc. In contrast to LEDs, these heads are equipped with actuators and moving parts which work in intense coordination. Romeo, iCUB, Simon, RoboThespian, MERTZ and KOBIAN RII, are some of the humanoids featuring kinematic head with moving eyelids, eyebrows, jaws and neck. In quest to manifest ultimate humanness, roboticists experimented with animatronic heads with flexible skin. Alice, Albert HUBO, Roman and Actroid are some of the examples of robots with animatronics head consisting of several DC motors and artificial skin made of special material called Frubber. Figure 3 shows three different kinds of humanoid heads with their ability to express emotions. Nevertheless such humanoid heads have a tendency to make users uncomfortable, as suggested in Mori's "Uncanny Valley" theory [6].

#### 2.1.1.2. Whole-body dynamics and control

The idea of substituting humans with surrogates for tasks like search and rescue in challenging scenarios has been prevailing for some time now [8]. With the introduction of socially interactive and socially assistive robots, designing humanoids to be autonomous has become inevitable. In a rich social setup, robots require high level of autonomy including extended physical mobility. Although robots are becoming more sophisticated both mechanically and emotionally, yet they are still far from achieving agile human-like manipulation and interaction, thus providing significant research potential in these areas. Dimensions of robot's body (i.e. height and weight etc.), Degree of Freedom (DoF), tactile sensors, number and flexibility of joints are the design factors that determine its mobility (i.e. walking, sitting, standing and turning etc.) and manipulation (i.e. reaching and grasping, pulling and pushing and holding etc.) capabilities. Whole-body control techniques [9, 10] have matured over the past few years enabling various humanoids to interact with their environment in a more robust manner. There is a steady transition of robot's actions in predictable contacts to unpredictable ones. Forums like DARPA Robotics Challenge (DRC)<sup>1</sup> , RoboCup2 and other international robotic challenges [11]

carry out human like tasks must be equipped accordingly; visual human likeness may not be of much importance in such cases as in the case of ATLAS and similar humanoids (Figure 2a). On the other hand those designed for interaction purposes must be more human like, with distinct facial expressions (e.g. Sophia) (Figure 2b) or with emotional speech capabilities (e.g.

With the aim to achieve a naturalistic embodiment, designers get inspiration from nature itself. Morphological design of natural looking social robots can be attributed to anthropomorphism. Based on their area of application, morphological inspirations for a robot's outlook can also be taken from zoomorphism (e.g. pets or creatures), caricature (e.g. animations or fictional characters) and functional expectations (e.g. assistive or service robots etc.). Nevertheless most social robots are intended to work with humans; thus the general notion is to give them a

human-like appearance. Therefore we will emphasize more on anthropomorphism.

Pepper) (Figure 2c).

Figure 2. Advance humanoids: (a) ATLAS, (b) Sophia, (c) Pepper.

6 Human-Robot Interaction - Theory and Application

<sup>1</sup> https://en.wikipedia.org/wiki/DARPA\_Robotics\_Challenge

<sup>2</sup> www.romela.org/robocup/

has its own challenges (noisy environments, multiple individuals talking, etc.), natural language processing has emerged as an important component of social robots which are expected to converse rather than simply accept keywords as commands. Such language models are important for the robot not only to understand what is being spoken but also to respond. Lee et al. [18] investigated the speech and language technologies involved in educational social robots and studied their impact in language learning. Brick and Scheutz [19] argue that robots must carry out their language processing incrementally with the ability to comprehend the context in order to meet the expectations of humans. Authors propose an interesting interaction engine (RISE) which incrementally processes the syntactic and semantic information. User modeling for effective natural language processing in long term human-robot interactions has

Socially Believable Robots

9

http://dx.doi.org/10.5772/intechopen.71375

As we try to make machines that look and behave like people, we need to equip them with perceptual abilities similar to that of humans. As user expectations exceed, a robot's perception must go beyond basic functionalities (e.g. localization, navigation or obstacle avoidance etc.). A key mechanism to achieve this is user modeling. Comparative studies of humans and robots can lead to new approaches [21–23]. The classical approach is to deliberately abstract computational instructions from physical realization of a human's cognitive system. Such robots that can perceive, infer and learn to mimic human behaviors are called cognitive robots. Intelligence, in a cognitive robot is the ability to transform sensed information into behavior. Human beings exhibit multitude of communicative signals while interacting. For a successful social interaction, a socially interactive robot should recognize the interaction roles, verbal and non-

verbal cues and situation; thus exhibit a considerable degree of "social intelligence".

Speech signals contain information about who is saying, what is being said and how it is being said. Context, tone, pitch and loudness all combine to convey information. Research regarding speech understanding in robotics include works like [24, 25], etc. In addition to vocalization, facial expressions, also give an insight into the intent of the social agent. Detection of human face and recognition of facial expressions is being incorporated in a socially believable robot. Cognitive empathy [26] is the phenomenon which models perception of emotions in robots. Gaze tracking [27] is another important aspect of perceiving the intentions of people while interaction, as it can indicate the focus of their attention. However gaze tracking involves detection of both face and eye orientation. Work is being carried out in this area but there are still numerous challenges that need to be addressed. Gestures and activity recognition [28], is also a promising area of research that can contribute to designing a socially intelligent robot.

Emotions play a significant role in human interaction; thus it was inevitable to induce emotions in socially interactive robots. The use of artificial emotions in social robots helps enhance believability and provides feedback to the users regarding the internal state of the robot, its goals and intentions. Artificial emotions [29], can also act as a control mechanism to understand robots perception of its surroundings. Numerous architectures have been proposed for

also been proposed [20].

2.3. Cognition and perception

2.4. Emotions and personality

Figure 3. Comparison of three humanoid faces based on emotion expression capabilities [7].

are providing a platform for innovative researches in this area. Nevertheless there is always room for further improvement especially in the out-of-routine challenging situations, multiple and diverse contacts etc. The concept of sight in robots is now possible due to various components like servo-motors, actuators, 2D or 3D cameras and embedded optical sensors. Computer vision techniques for object recognition, human gestures, gaze and speaker tracking and collision or obstacle avoidance mimics sense of sight for the humanoids [12]. Distant communication with a robot using voice in an unconstrained environment is a highly challenging task. Methods to improve the auditory and speech recognition of a robot are being given much attention by the researchers [13–15].

#### 2.2. Speech and linguistics

Speech is the most effective and natural mode of communication and interaction. From the view point of social robots, not only the robots need to be equipped with state-of-the-art automatic speech recognition (ASR) software [14], language models for interaction [16, 17] are also required to make semantic sense of what is being communicated to the robot. While ASR has its own challenges (noisy environments, multiple individuals talking, etc.), natural language processing has emerged as an important component of social robots which are expected to converse rather than simply accept keywords as commands. Such language models are important for the robot not only to understand what is being spoken but also to respond. Lee et al. [18] investigated the speech and language technologies involved in educational social robots and studied their impact in language learning. Brick and Scheutz [19] argue that robots must carry out their language processing incrementally with the ability to comprehend the context in order to meet the expectations of humans. Authors propose an interesting interaction engine (RISE) which incrementally processes the syntactic and semantic information. User modeling for effective natural language processing in long term human-robot interactions has also been proposed [20].
