Towards A Robot-Assisted Autism Diagnostic Protocol: Modelling and Assessment with POMDP

The existing procedures for Autism Spectrum Disorder (ASD) diagnosis are often time consuming and tiresome both for highly-trained human evaluators and children. In addition, prospective human evaluators need to undergo a rigorous and lengthy training process that may not be accessible or affordable to all interested individuals. Hence, this paper proposes a framework for robot-assisted ASD evaluation based on Partially Observable Markov Decision Process (POMDP) modelling. POMDP is broadly used for modelling optimal sequential decision making tasks under uncertainty. Spurred by the widely accepted Autism Diagnostic Observation Schedule (ADOS), we start off with emulating ADOS. In other words, our POMDP model explicitly takes into account the ADOS stratification into several modules, ongoing task informativeness and robotic sensor deficiencies. Relying only on imperfect sensor observations, the robot provides an assessment of the child’s ASD-relevant functioning level (which is partially observable) within a particular task. Finally, we demonstrate that the proposed POMDP framework provides fine-grained outcome quantification, which could also increase the appeal of robot-assisted diagnostic protocols in the future.