Technical Writing

Humanoid that Interacts with People

As number of Hollywood fantasies have portrayed, the future of AI promises there will be lots of robots in our everyday lives. Some, perhaps many, of them could look like human and behave like people (humanoid). It also desirable for them to produce as little as possible cognitive load on humans, which means they must have an ability to understand human social conventions i.e. conversation, gesture, etc (interactive).

Those two key points, humanoid and interactive, have lead to the development of Kismet, robot with humanoid form, which intended to act humanly as people do in the world. Kismet built with social interactions ability, which themselves based on eye movements, facial expressions, and voice prosody detection.

The Kismet eye movements are the bases for their social interactions. Even an ordinary observer can understand what the robots are paying attention to by the direction of their gaze. Likewise, the robot can understand what a person is paying attention to by the direction of the person’s gaze. The visual attention system makes it completely intuitive for ordinary people to direct the robot’s eyes to some particular objects. If an object was brought into the field of view of the robot, and moved to the desired position, the robot will smoothly pursue it, paying attention to where the object goes. Of course the robots are not able to do visualization as good as human visual system, but estimation with 4 or 5 degree accuracy is enough for social interaction. Kismet is also able to detect human faces through a variety of methods and estimate the gaze direction of a person by determining the direction their eyes are pointing. By manipulating the weighting of the visual system, it will create different visual sign. As a result, Kismet has high-level human-like behaviors, such as dialogue turn-taking and making eye contact. Indirectly, these high-level behaviors regulate social interaction.

Moreover, Kismet expresses its internal emotional state through facial expressions and prosody in its voice. So, for instance, when someone comes very close to Kismet or waves something very quickly nears its face, Kismet becomes fear. That emotional state is reflected in its posture; it draws back. Thus, Kismet, indirectly through its emotional system and its expression, is able to manipulate people, just as humans unconsciously manipulate each other.

Kismet is also able to detect basic patterns and rhythms in the voices of people and classify them into praising, prohibiting, comforting, or interest, four basic prosodic signals used in almost all human cultures. Nevertheless, the robot could understand only the prosody not the actual words said.

--

Morenvino M (135 04 034)

Nicolas A (135 04 109)