SPOON, ENSTA and Strate PhD project
Behavioral Development Model for Smart Creatures in Contextual Social and Collective Interactions
Description
The challenge of this PhD is to determine the elements which allow a robot to adapt to the societal context in which it evolves, where it can be accepted and adopted. A truly personalized social robot must take into account the likes, dislikes, preferences, and limitations of a human and adapt its behavior accordingly. In this PhD, we will develop a set of tools and models that we will integrate into the artificial creatures - the SPooNies - of our industrial partner, the startup SPooN, via a system called the “Creature Development Kit" (CDK).
The Creature Development Kit for a contextualized interaction
The CDK is a set of reflections but also a set of tools, based on large scale experiments in order to better address everyday needs and to understand what form, in a broad sense, a creature should take in different contexts. We want to continue implementing adaptive architecture of an interactive character to offer the user means to configure this interactive character in the most natural and accessible way possible.
With this CDK, the user can interact by text but also vocally and by other modalities to determine the characteristics and behaviors of these characters. Thus, through this interaction, the user can describe the appearance, character, behavior and interactive modes of the SPooNy for a given location. The algorithms developed by SPooN transform these interactions into a semantic space, generating a dictionary of the world. Thus the SPooNy is in charge of "understanding" this semantic description of the world with its current capabilities (sensors, skills, utilities, etc.). It will be able to do so by interacting with its environment (in different scenarios). As a result of these exchanges, the SPooNy's dictionary of the world will grow and therefore become better adapted to the context and to the people with whom it interacts.
Towards the convergence of objectives
The goal is to converge all areas of work and the tools used towards a single driving idea: a personalized, natural interaction. This concept really links the research and industrial approaches. Indeed, the more SPooNy manages to blend into an environment, while giving users a feeling of customization or adaptation, the easier it will appeal to a wide market. Moreover, the interest of these models and the study of their interactions with users depends greatly on these elements of customization. One of the strengths of this PhD is that as it is directly intended for integration in a pre-existing system, these models will be intended for a concrete product; which will lead us to seek a minimalist form of efficiency.
The importance of the human-machine relationship
This integration into a pre-existing context also modifies the approach of this PhD into another specific field. One of SPooN's ambitions is to redefine the relationship between humans and machines for more natural and constructive interaction. The very term "robot", because of all the negative connotations it evokes, could almost become an obstacle. One thinks in particular of science fiction, even if the etymology of the word "robot" is connoted since it comes from the theatrical performance RUR by Karel Čapek; the term robota meaning "chore" or "serfdom" in Czech. Thus, to avoid creating barriers, it was decided that SPooNies and all robots developed by SPooN would be called "artificial creatures".
The strength of context to deal with the notion of locality
Understanding what people expect from a robot and what it entails will be a real challenge. This is where the study of diverse fields such as animation theory or the study of intermediate objects are important. This will require communicating with numerous and diverse individuals in different contexts. For this reason, the hybrid aspect of a CIFRE PhD is an immense advantage. It allows the PhD student to express himself or herself in multiple environments by combining research and industry. As two academic laboratories are involved, this is all the truer here. It also highlights the need for experimentation, and a flexible and organized methodology. Another asset is the multidisciplinarity of the team, with the PhD student and his supervisors coming from different specialisations, which will facilitate combining the various study objects. This aspect is further reinforced as this PhD is hosted by Robotics by Design Lab alongside three other thematically related theses with different profiles and specialisations.
Supervisors
There will be many questions which the PhD student will try to answer with the help of the thesis director Adriana Tapus, professor at ENSTA Paris, under the supervision of the co-director, Ioana Ocnarescu representing Strate and Robotics by Design and the industrial Ugo Cupcic from Spoon.