The interaction of social robot with autistic children greatly improves their therapy but requires human intervention to control the robot’s behaviour. In this work, a cognitive architecture based on a multi-sensors approach has been developed to drive a Nao robot autonomously. After the definition of criteria, a review of the literature has allowed to select the "ethological model of drives and needs" as candidate architecture. This abstract model has been adapted for robot interaction and has been implemented in a software based system. Three off-the-shelf sensors; namely the Kinect 2.0, a Webcam and Nao’s microphone, have their signals processed by dedicated software to capture respectively the body position, the mood and the answers of the person interacting with Nao. These high-level and detailed information are used as input by the cognitive system to drive a set of output actions based on the needs defined in the system. These output actions control the behaviour of the robot. The system developed has been tested in real conditions with 10 children, proving that the robot is able to interact autonomously. Nevertheless, an additional test with 16 other children has not clearly demonstrated that the children have the same perception of the robot when it is controlled by a human or by the cognitive system. Possible improvements have been identified to increase the emotiveness of the Nao robot. Ways to increase the stability, coherence and scalability of the cognitive architecture have also been explored; setting the grounds for a Layered Ethologically Inspired Architecture (LEIA).
Datum prijs | 2014 |
---|
Originele taal | English |
---|
Prijsuitreikende instantie | - Vrije Universiteit Brussel
- Université libre de Bruxelles
|
---|
Begeleider | Bram Vanderborght (Promotor) |
---|
An autonomous cognitive architecture for robot therapy
De Beir, A. ((PhD) Student). 2014
Scriptie/Masterproef: Master's Thesis