Virtual Reality and Robots to study Knowledge and Embodiment role in Attention and Anticipation control in Dynamic and Social Environments
Many competitive and collaborative social interactions rely on the ability to anticipate others’ actions. This ability poses computational and sensory resources   under heavy pressure and is a main limit in current Human Robot Interaction systems.
For example, a boxer during a fight  has to anticipate next move of the adversary for which many clues are important, e.g. where is him looking, on which foot is him putting more weight, what muscles are more tense, what is their mutual position and their position in the ring.
All these important information sources are not simultaneously accessible and how the human brain is able to efficiently select them is not clear.
Most of the current gaze control models are mostly based on spatially and temporally local features of the visual input, however a key role can be played by the observer knowledge of the environment dynamics which may offer a more general solution for predictively observing the environment, e.g. less prone to fall for a feint. Another important factor can be the non-linear structure of retina which may explain the perceived predictive nature of many eye movements.
The student will study the visual exploration strategies of human subjects immersed in virtual environments populated by one or more virtual humanoid agents interacting with the environment and create different virtual human observers using the collected data and different hypotheses on how knowledge and embodiment can affect gaze control. The role of knowledge, retinal structure and visual features will be investigated varying the virtual reality experience of the subjects. The models could be validated through reverse Turing test in the virtual environment.

Leave a Reply