Assistive Robotics for Autism Therapy: Programming Robots to Express Emotions

Abstract

Autism, a neural development disorder, occurs in different degrees and forms. People with autism have great difficulty making sense of the social world around them. They also have difficulty expressing their feeling and thoughts verbally and non-verbally. These characteristics set them apart socially [24]. There is no cure at present for autism, but therapists, doctors, and teachers are trying their best to identify therapies that reduce the negative effects of the disorder. Robot technologies can potentially be used in autism therapy programs such as Applied Behavior Analysis (ABA), Play Therapy and Prosodic Therapy. Research has shown that children with autism have great affinity towards robots. However, since each autistic child is different and there is no generalized characteristic behavior of autism, robots must be tuned to the specific conditions in which they are used. Situation-specific robot behaviors need to be developed. Since, teachers, parents and therapists are not technology experts, programming interfaces must be developed that allow them to effectively choreograph robot behaviors and skills for interactions with specific autistic children. Given that emotions play a critical role in human-robot interactions, we focus in this research on the abilities of non-technology experts to program robots to express emotions. In so doing, we seek to identify the programming interface characteristics that are necessary to allow non-technology experts to choreograph these emotional behaviors effectively into robots. The research will help us to learn whether robots without facial expressions can be used for expressing emotions and to determine how effectively end users can choreograph robot emotions using the existing interfaces. We also study the characteristic of each emotion with respect to robot body posture and sound, and investigate how the recognition rate of the choreographed emotion by the participants can be increased. Results are established using a series of user studies involving 54 participants, who program and identify emotions in Nao and Pleo robots. Initial results indicated that verbal expressions were much more salient to users than non-verbal expressions. Follow-up user studies involving 54 participants were then conducted to further investigate systems for creating better verbal expressions in robot behaviors using Nao. Results indicated that first creating body language for a robot and recording sound while playing the body movement simultaneously helps to create a better verbal expression. These latter user studies also reinforce our previous findings that verbal communication is the most effective method for people to program robots to express identifiable emotions.

Date of Award	2011
Original language	American English
Supervisor	Jacob Crandall (Supervisor)