Current students


ESPOSITI FEDERICOCycle: XXXVII

Section: Computer Science and Engineering
Advisor: BONARINI ANDREA
Tutor: AMIGONI FRANCESCO

Major Research topic:
VIRTUAL SOCIAL INTERACTION OF NON-HUMANOID AVATARS THOUGH EMBODIMENT

Abstract:
The Metaverse can allow us to embody, in principle, any body we want, both digital(VR) and real(robotics, telepresence). But there is very little research at the moment on inhabiting avatars that are not humanoid, thus going beyond human-like features and movement capabilities.
My research addresses these limitations, exploring the basic features of avatar embodiment with non-humanoid avatars, with the final objective of obtaining meaningful interactions between users though their avatars. After all, interaction is the key feature of a Metaverse.
To tackle this problem, two aspects will be considered: avatar design, free from humanoid structure and thus characterised by the essence of movement features, and the hardware design for the user, researching the possibility to improve the current typical VR setup with more control and feedback components. 
One of the elements of embodiment is the sense of “agency” that the user needs to feel when the avatar performs actions. In order to test this we thus need an environment where actions can be taken that have an effect. In this research I will explore simple minimalistic environments focused of social interactions between users. 
The main goal of this research is to move towards a type of metaverse that is not being explored fully at the present time: the Metaverse of physical avatars in the real world, where mechanical creatures interact while remotely inhabited by human users. 
Research on embodiment and presence beyond the humanoid shape is very relevant at this time. First of all, from a psychological and psychosocial point of view, the results of such a research will shed more light on the essence of what we perceive as a “body” and “communication”, beyond the shapes to which we are accustomed to. A decade of animation characters have already demonstrated that, given some conditions, humanoid bodies are not needed for an entity to be considered “alive” (both as autonomous and social). So the next natural step is to explore how to inhabit such bodies and use them to interact with others.  Secondly, this research is relevant as it will contribute to the research on VR and metaverse technologies. This field is being explored consistently at this time, but not as much from this point of view. Contributions will span from the exploration of variations to the common control and feedback mechanisms in VR, to the design of embodiment-centric avatars, to system design and implementation for fast and efficient telepresence. Lastly, this research will contribute to the field of robotics in two ways. One, when designing movements and behaviours for autonomous robots, usually these are firstly tested in Wizard of Oz fashion. This process is still limited to the control and feedback design, which is usually very simple and limit the exploration capacity of this texting phase. Second, more importantly still, as in this research systems for embodying any-shape physical robots will be developed, it will also provide a dataset of the behaviours of users when inhabiting such robots, thus the type of behaviours that these robots will express. This will be an invaluable instrument for research on automatic robot behaviour, as it could be leveraged by machine learning tools to extract common relevant features or directly learned from with off-policy RL techniques. Possible applications will thus range from entertainment, to research fields as an effective tool, to the industry: freeing designers from the humanoid shape while maximising the user capacity to inhabit any-shape physical robots, teleoperation tasks will become more efficient even for very specific activities that require specific robotic shapes, such as quadcopters or wheeled robots.