DESCRIPTION

Artificial Intelligence and Cognitive Models for Symmetric Human-Robot Interaction in the field of assistive robotics

COMPANION-CM is a synergistic project funded by “Comunidad de Madrid”, with reference Y2020/NMT-6660, in the framework of the 2020 call for synergistic R&D interdisciplinary projects in new and emerging scientific areas at the frontier of science.

Every day the number of people with reduced mobility, cognitive problems and/or limited language abilities who face a daily routine in solitude, either at home or in hospital settings, grows. Hence, maintaining the necessary contact with them has become a great concern for the health community (a situation that is currently aggravated by covid19). In order to alleviate this need, COMPANION-CM
project aims to develop assistive robots with a more empathic and symmetric interaction capacity, personalizing Human-Robot interaction (HRI) through Artificial Intelligence (AI) techniques.

The main goal of the project is to build a training framework integrated over the humanoid robot TEO
(developed at UC3M) so that its AI learns during its interaction with humans and consequently personalizes the HRI to them. Such a learning system will be supported by the development of a series of disruptive enabling technologies in the field of AI. Finally, the framework will be evaluated in three demonstration scenarios: a) assistive robots in a residential setting, b) companion robots for hospitalized patients; and c) rehabilitation robots for patients with cognitive problems.

Objectives

General Objective

The main objective of the Project is the development of an integrated training framework embedded on the humanoid robot TEO, developed at UC3M, so that its artificial intelligence learns during its interaction with the human and personalizes the HRI between them. The learning system will be supported by the development of a series of disruptive enabling technologies in the field of AI and will incorporate characteristics of human-to-human communication, such as the “recipient desing” (model of the interlocutor, prediction of intentions, mood, etc.), and mechanisms to improve the interpretability of the robot’s behavior by the human.

The project aims its application in the field of a more social and reliable assistive robotics for a group of patients with reduced mobility, cognitive problems and/or limited language ability, who are assisted by a robot in residential or hospital environments.

Specific Objectives

Scientific and technological

OBJ1. Transfering “recipient design” capability from human-human communication to HRI through computational modeling.

OBJ2. Designing interpretable and explainable models of human behavior from the analysis of multimodal evidence (sensors) and using Deep Generative Neural Networks improved by the relevant cognitive ingredients.

OBJ3. Providing computational models to decode human intention from physiological signals (wearable sensors) and to anticipate their actions.

OBJ4. Developing the robot an intelligent manipulation strategy using non-verbal commands.

OBJ5. Develop a learning framework that provides a more symmetric and bidirectional personalized HRI, using cognitive models of the internal state of the interlocutor, and establishing a bidirectional communication with the human.

OBJ6. Deploying three demonstrators in the field of assistive robotics to evaluate the scientific and technological developments of the project: 1) residential scenario; 2) companion robots for patients in hospital settings; and 3) rehabilitation robots for patients with motor and cognitive impairments.

Transversal

TO1. Creating an interdisciplinary community and an innovation ecosystem oriented to the project’s objective that integrates researchers, technology companies and hospitals.

TO2. Promoting collaboration between researchers from different fields and the training of young research talents with multidisciplinary knowledge.