Monday, 6 February 2012

The Art of Designing Robot Faces

Robots are becoming available in a wide variety of roles. A recent report by the UN Economic Commission for Europe and the International Federation of Robotics predicts that 4.1 million robots will be working in homes by the end of 2007. The implication is that, as they crawl, roll and walk out of the laboratory and into the ‘real’ world, that people in the real world will be using them - families, soldiers, nurses, teachers. These users will more than likely not have a background in engineering nor care about the intricacies of the control algorithm in their robot. To them it will be a tool in the same way as a DVD player or a PC. However robots differ significantly from most consumer electronics in two respects: Robots are often designed (and expected) to use human communication modalities, for instance speech and hearing in place of LED displays and buttons. This is sometimes because these modalities are implied by the robots’ anthropomorphic design and sometimes for practical reasons - robots are usually mobile and even a remote control may be of limited practical use. Due to their embodiment, robots have the capability to supply rich feedback in many forms: anthropomorphic ones such as speech, gestures and body language and ‘artificial’ ones such as lights and music. Current consumer robots such as the Sony AIBO use a combination of both. Using real-time communication a robot can engage the user in active social interaction, and importantly even instigate interaction. In contrast, most consumer electronics are passive; that is, there is only interaction when instigated by a human, and that interaction is largely in one direction, from the human to the machine.

A user study of people’s expectations of a robot companion indicated that a large proportion of the participants in the test were in favour of a robot companion, especially one that could communicate in a humanlike way.

DiSalvo et al. performed a study into how facial features and dimensions affect the perception of robot heads as humanlike. Factors that increased the perceived humanness of a robot head were a ‘portrait’ aspect ratio (i.e. the head is taller than it is wide), the presence of multiple facial features and specifically the presence of nose, mouth and eyelids. Heads with a landscape aspect ratio and minimal features were seen as robotic. They suggest that robot head design should balance three considerations: ‘human-ness’ (for intuitive social interaction), ‘robot-ness’ (to manage users’ expectations of the robot’s cognitive abilities) and ‘productness’ (so the user sees the robot as an appliance). The idea of designing a robot to be perceived as a consumer item is noteworthy for the fact that people’s a priori knowledge of electronic devices can be utilised in avoiding the uncanny valley; the implication is that the robot is non-threatening and under the user’s control. To fulfill their design criteria they present six suggestions: a robot should have a wide head, features that dominate the face, detailed eyes, four or more features, skin or some kind of covering and an organic, curved form.

Role Suitability.
Goetz et al. introduced the ‘matching hypothesis’, i.e. that machine-like features are seen as more suitable for investigative and security (i.e. authoritarian) situations, whereas those with a human appearance are preferred for artistic, service and social roles.

Appearance has a great effect on the perceived personality and emotional capabilities of a robot. Some robots, especially those sold as toys or for use in the home, draw on the long history of doll design and use abstracted or exaggerated infant-style features such as a round, symmetrical face and large eyes. This aesthetic evokes protective instincts in the user thus circumventing the uncanny valley. Woods, Dautenhahn and Schulz conducted a study in which children were asked to rate robots in terms of several personality traits according to their appearance. The robots were classified as having machine, animal, animal-machine, human-machine and human appearance. Results indicated that the overall combination of physical attributes was important, with animal-like robots rated as happy, purely mechanistic robots rated as aggressive and angry, human-machine robots rated as friendly and human-like robots rated as aggressive. Interestingly this result supports the uncanny valley as human-machine robots were rated more positively than realistic human-like ones. The children were also asked to assign each robot a gender, and it was found that those perceived as female were often associated with positive personality traits.

Sensory and interaction capabilities.
Users are likely to make an initial assessment of the robot’s interaction capabilities based on appearance. An ultra-realistic humanoid with limited movement or interaction abilities is likely to disappoint; a simple-looking robot that turns out to have complex interactive behaviour will surpass expectations

The consistency of the design of a robot might also be a consideration in managing user perceptions. A robot with mismatched features, for example a realistic ‘skin’ covered head with exposed mechanical limbs, may appear more uncanny than one that is aesthetically harmonious.

Humans are extremely sensitive to the particular pattern of features that form a face. [...] We cannot help but see faces in everything [...] Faces can be abstracted or simplified by a huge degree and still remain recognisable, a useful characteristic for comic and caricature artists - and robot designers. Minimal features or dimensional relationships are all that is required to suggest a face, and our brains ‘fill in the gaps’.

Robot faces that mimic the prototypical dimensions of human faces are likely to stimulate the face processing system that we also apply to other people. One might thus speculate that a more ‘generic’/iconic face of a robot affords more scope for people to identify with, since it allows them to project upon it face representations acquired during their lifetime.

Faces are the focal point of any humanoid robot, but most suffer from some, or all, of the following problems: they are hard to make look realistic, and even if they do the illusion is often shattered once they move; they are complex, requiring many degrees of freedom (DOFs); they are expensive to make and maintain, and they are arguably the part of the robot most likely to pull the rest into the uncanny valley. Furthermore it could be argued that the feedback provided by a face can be more cheaply presented using some other modality such as LEDs. So the question is: are faces worth the trouble? Quite apart from the fact that they are by definition part of a humanoid robot, there are several good reasons for their use:

1. Expressions are a universally-used feedback mechanism and are instantaneously understood by an interaction partner. A red LED could be used to represent happiness, but the association has to be learnt and processed, and the colour itself might come with cultural connotations (for instance red symbolises good fortune in China, but danger in the UK). A smile has a less ambiguous and more immediate emotive impact.
2. A face gives the user an understood focal point for interaction. A face affords interaction.
3. A face can present visual cues to help the user understand the robot’s capabilities, forming an unspoken social contract between human and machine. A very mechanistic appearance may lead to confusion over communication modalities (“Do I talk to it? Will it understand me? How does it hear?”), whereas clearly-presented communicative features will encourage intuitive interaction. In addition the design of the face can give clues as to the ability level of the robot; a two-year old face implies two-year old cognitive and manipulative abilities.
4. Variable expressions can assist the robot in its role; for instance a face might allow a security robot to look friendly or intimidating as required, or allow a toy robot to look cute or express surprise in interaction games.

Having established that faces can be useful, how should they look? Despite the enormous variety in real human faces, most people are intuitively aware when something looks unusual. Cartoons on the other hand, using merely representations of faces, can cover a far larger aesthetic range. In his book Understanding Comics, Scott McCloud introduces a triangular design space for cartoon faces.

The left apex is realistic, i.e. a perfect representation of reality, for example a photograph, or realistic art such as that by Ingres. Travelling to the right faces become more iconic, that is, the details of the face are stripped away to emphasise the expressive features; emoticons such as :) are a perfect example in the 21st century zeitgeist. The simplification has two effects. Firstly it allows us to amplify the meaning of the face, and to concentrate on the message rather than the medium. Secondly the more iconic a face appears the more people it can represent. Dautenhahn points out that iconography can aid the believability of a cartoon character. We are more likely to identify with Charlie Brown than we are with Marilyn Monroe, as a realistic or known face can only represent a limited set of people whereas the iconic representation has a much broader range - to the extent of allowing us to project some of ourselves onto the character. Towards the top apex representations become abstract, where the focus of attention moves from the meaning of the representation to the representation itself.

As one moves in the design space of the faces from realism towards iconicity, a human is more able to identify with the face, and the distinction between other and self becomes less and less pronounced. In theory then the less detailed a face becomes the more characters it can represent. In addition, the more iconic a face is the more we are able to project our own experiences and emotions on it. Why is this? McCloud argues that the mental image we have of our own face is a subjective, iconic one. In contrast, the way we see others around us is objective and fully realistic. Hence the iconic face can represent us, and the detailed face represents somebody else.

Excerpts from:
‘The Art of Designing Robot Faces’by
Mike Blow,
Kerstin Dautenhahn,
Andrew Appleby,
Chrystopher L. Nehaniv,
David Lee

No comments:

Post a Comment