Towards Enhancing the Multimodal Interaction of a Social Robot to Assist Children with Autism in Emotion Regulation

Assistive robots are expected to become ubiquitous by transforming everyday life and are expected to be widely used in healthcare therapies. SARs (Socially Assistive Robots) are a class of robots that are at an intersection between the class of assistive robots and that of interactive social robots. SARs are being explored to assist in the diagnosis and treatment of children with ASD (Autism Spectrum Disorder). A SAR called EVA has been used to assist non-pharmacological interventions based on verbal, non-verbal communication and social interaction. The EVA robot can currently speak, listen and express emotions through looking. Towards offering immersive therapies for autistic children, this work enhances EVA's capabilities to recognize user emotions through facial expression recognition and also to create light sensory effects in order to make the therapy more attractive to users. A therapy session was developed through a serious game where the child should recognize the robot's emotions. During the game, EVA recognizes the child's facial expression to check his/her learning progress. We invited a neurotypical 6-year-old child to play the game, with the consent of her parents, and recorded videos of the game session. Those videos were evaluated by 48 expert physicians and psychologists in therapies for ASD using the Technology Acceptance Model (TAM). They considered our work useful and agreed it would help them doing their job more effectively.