The Optimal Social Assistive Robot for Autism Spectrum Disorder Interventions

Abstract

People with Autism Spectrum Disorder (ASD) struggle with deficits in social communication and interaction. To help with these deficits, early, intensive and behavioural interventions are recommended. People with ASD have shown a higher degree of engagement through the interaction with robots than through interactions with other humans. Robot interventions for ASD have therefore been one of the first applications explored in the field of Social Assistive Robots (SAR). However, these robots are still semi-autonomous and need to be controlled by professionals at clinics which can be disadvantageous to the learning abilities of the child. The solution proposed in this thesis is to make SARs autonomous with the use of Artificial Intelligence (AI). Making a SAR autonomous to be used for ASD interventions in a home setting influences the current design of a SAR. The current intervention types and designs of SARs will therefore be discussed and what needs to be applied to make a SAR autonomous. Lastly, a SAR will need social intelligence to work during an intervention. The requirements to create social intelligence with the use of AI will be discussed and the differences between general social intelligence and social intelligence specifically for ASD interventions will be highlighted.