Robots help autistic children
DREAM is a major new €8m project major new project funded by the European Commission and coordinated by the University of Skövde. The goal is to deliver the next generation robot-enhanced therapy (RET) for children with autism spectrum disorder (ASD), developing clinical interactive capacities for social robots that can operate autonomously for limited periods under the supervision of a psychotherapist.
In clinical interventions, skill transfer from therapist to children with ASD is helped by the use of expressive artefacts such as puppets and animated characters. Well-designed robotic agents have proven to be particularly effective and are becoming an increasingly important tool for mediating between therapists and ASD children in what is known as robot-assisted therapy (RAT). However, therapeutic interventions require significant human resources over extended periods. Consequently, to make a significant difference, therapeutic robots need to have a greater degree of autonomy than current remote-controlled systems. Furthermore, they have to act on more than just the child’s directly-observable movements because emotions and intentions are even more important for selecting effective therapeutic responses.
The next generation of RAT, which we refer to as robot-enhanced therapy (RET), will be able to infer the ASD children’s psychological disposition and assess their behaviour in order to select therapeutic actions. Since children require therapy tailored to individual needs RET robots will provide this too. The DREAM robot will also function as a diagnostic tool by collecting clinical data on the patient. It will operate under strict ethical rules and the DREAM project will provide policy guidelines to govern ethically-compliant deployment of supervised autonomy RET.
The core of the DREAM RET robot is its cognitive model which interprets sensory data (body movement and emotion appearance cues), uses these percepts to assess the child’s behaviour by learning to map them to therapist-specified behavioural classes, and then learns to map these child behaviours to appropriate therapist-specified robot actions.