This research designed and applied 24 multi-purposeful robot activities of varying social mediation levels in a multiple-session experiment with 34 children of diverse autistic characteristics in a rehabilitation setting. This paper explores what type of robot activities can meet individual needs to bring more socio-behavioral progress and juxtaposes child characteristics to identify behavioral outcomes in each activity. This knowledge would help us to respond to the question of what activity types suit specific subgroups of Autism Spectrum Disorder (ASD). Our data analysis included coding 48.5 hours of video data for a total of 14 measures to fully capture children's activity-based socio-emotional outcomes. Overall, the activities on varying social mediation levels brought more or less positive social outcomes to all children. However, children showed some different behavioral outcomes as mediated by core autism-related and age-specific characteristics. This study provides in-depth accounts of what might be helpful in designing and applying multi-purposeful activities responsive to the diverse needs of children.

Social robots are increasingly being used as a medi-ator in robot-assisted autism therapy to improve children's social and cognitive skills. Engagement is one of the key measurements used to evaluate the therapeutic interventions' effect on children. While “engagement” is broadly used, it has been challenging to find a consensus about its definition in the community. With this paper, we explore the use of a data-driven approach to investigate the extent to which a model on engagement built on one dataset transfers to another. We utilized two publicly available datasets of engagement recognition, namely PInSoRo and Qamqor datasets, with an attempt to achieve a higher accuracy taking into account the transferred knowledge. The accuracy of 83.18% was obtained on the PInSoRo dataset of child-child interactions with face and body keypoints. We have used the methodology of transfer learning to improve the classification accuracy on the Qamqor dataset. The best result obtained is a 71.89% accuracy on the Qamqor dataset. This suggests that more data with similar keypoints is needed to achieve better accuracy when utilizing transfer learning from one dataset to another dataset.

Research has recognized the importance of individ-ual differences of children with Autism Spectrum Disorder (ASD) that require interventions to meet their heterogeneous needs. This relatively large-scale study investigates a robot-assisted autism therapy (RAAT) with 34 children with diverse forms of ASD and Attention Deficit Hyperactivity Disorder (ADHD). We conducted a multi-session study with multi-purposeful activities targeting the socio-emotional abilities of children in a rehabilitation setting. We found a number of quantitative results suggesting various autism-related and demographic differences such as diverse forms of ASD, co-occurrence of ADHD, verbal skills, and age groups. The main findings are: 1) severity of ASD forms may not predict intervention outcomes but instead the co-occurrence of ADHD with LFA diagnosis may negatively impact social smiling; 2) verbal children were more generally engaged and less aggressive with the robot than non-verbal children whose curiosity rose over sessions; and 3) younger children (3.4 y.o.) showed more affection, while older children (7–12 y.o.) were better engaged through speaking more words and having longer engagement and eye contact with the robot.

Social robots deployed in the therapy of autism is a promising and important research domain. Recently, an increasing amount of work is being conducted utilizing a social robot as a mediator between a therapist and a child with autism. Being able to evaluate how engaged a child is both offline and in real-time would improve the quality of the provided robot-assisted intervention and also provide objective metrics for later analysis by the therapist. The state-of-the-art engagement recognition is challenged by the diverse styles of expressing engagement by this vulnerable population group. To this end, this PhD project aims to explore how transfer learning can improve the recognition accuracy of children's engagement with the robot or another human. We will utilize four publicly available multi-modal datasets to discover a suitable feature representation of engagement during various types of activities with the robot.