Development of Socially Assistive Robots For Children With Autism Spectrum Disorders

Abstract

This paper describes the design and implementation of a robotic system for interaction with children with autism spectrum disorders (ASD). The system consists of a robot and supporting infrastructure, including a “smart room”. The intent of this work is to explore and study the design of a therapeutic, minimally-restrictive environment that enables free-form human-human and human-robot interaction. Our primary design goals include: 1) using minimal structure with the participants in order to elicit natural behavior; 2) increasing a child’s social interactions while utilizing minimal human-operated technology; and 3) facilitating human-robot interaction while requiring the child to wear no more than the minimum needed for effective signal detection. The robot system implemented in this study uses non-invasive methods for sensing and interpreting the child's behavior in order to interact with a child in free-form play while eliciting social behavior from the child. We present results from two feasibility studies with 12 children with ASD in order to validate the effectiveness of the robot system. We also discuss recommendations for the use of robot technology in ASD research settings. Autism is a biologically-based disorder affecting social-communicative development (Dawson et al. 2002) characterized by symptoms often observable by 24 months of age, although the symptoms may present earlier (Stone, Coonrod, & Ousley, 1997), and currently affects between 1 in 150 children (CDC 2007). One of the primary social-communication disturbances characteristic of autism is an impairment in social attention and/or initiation of social behavior that manifests as a lack of spontaneous seeking to share experience, enjoyment, interests or achievements with other people (Mundy & Crowson, 1997). Children with autism have difficulties with both social interactions and verbal communication skills (Mundy and Crowson 1997). The paucity of these skills presents challenges in the implementation of existing behavioral interventions, since these interventions include reinforcing naturally occurring social behavior (Koegel, et al., 2003; Mundy, et al., 1997). To date, the focus of much early intervention research in the autism literature has been to develop methods addressing early symptomatic impairments in joint attention (Kasari et al., 2006; Jones et al., 2006; Mundy & Crowson, 1997). Research in this area suggests that the early improvement of joint attention development in autism may ameliorate a negative developmental cascade of language and social-cognitive impairments that ensue when this pivotal arena of developmental disturbance is not effectively addressed (Kasari et 2006). However, interventions are still challenging to implement and do not result in improvements for all children with autism. One explanation for this may be that children with autism are less intrinsically motivated to initiate social behaviors and therefore do not engage in treatment activities. Thus, one important component of an effective intervention for children with autism is a child-focused approach whereby adults identify the child’s intrinsic interests in order to gradually engage them in interactions, which will maximize the child’s attention (Kasari et al. 2006; Siller & Sigman, 2002). Socially Assistive Robotics (SAR) seeks to provide assistance to users through social interaction (FeilSeifer and Mataric' 2005), and is being studied for therapeutic use with children with ASD. It has been observed that children with ASD interact with robots differently than with people or toys and may show intrinsic interest in such machines. This intrinsic interest could be applied as a robot augmentation for an intervention for children with ASD. Werry, et al., (2001) conducted an experiment in which children with ASD and their teacher interacted with a robot and a non-animated toy. Behavioral coding of the experimental data demonstrated an increase in physical contact and eye gaz