Robot-mediated telerehabilitation system using internet of robotics things (IoRT) for autism intervention

The convergence of Fourth Industrial Revolution (4IR) technologies has catalysed growth in assistive technology for special needs, including robot-mediated telerehabilitation systems for autism intervention. These systems aim to bridge the distance between Children with Autism (CwA) and therapists using robots. However, the implementation of telerehabilitation technologies in the systems is still in its infancy, leading to considerable inefficiencies and restricting the deployment of advanced functionalities of humanoid robots such as NAO. Thus, this study aims to develop an Internet of Robotics Things (IoRT)-based robot-mediated telerehabilitation system for autism intervention and investigate its feasibility, functionality, and performance. In this system, a CwA interacts with a robot, assisted by a guardian, while connected to a far-apart therapist via IoRT architecture. IoRT, a multi-layer architecture (network, perception, middleware, and application) was established with infrastructures such as NAO robot, Eclipse Mosquitto Message Queuing Telemetry Transport (MQTT) broker, MySQL database, Node-RED programming tools, Linux server and DigitalOcean cloud hosting. Here, the network layer performance test was prioritised, followed by a feasibility test confirming basic data transmission. Then, comprehensive user-oriented developments for perception, middleware, and application layers were conducted and evaluated using performance or functionality tests, followed by real-world experiments on the integrated system. In terms of results, the network layer's performance tests measured by JMeter showed low latency for MQTT publish-subscribe by achieving 90.0 % percentile below 300 ms, with zero error, affirming its efficiency and reliability. The feasibility test confirmed the successful data transmission across all layers, validating the architecture for a basic intervention module. The perception layer's eye gaze implementation showed good agreement with frame-by-frame video analysis, with acceptable limits in Bland-Altman analysis. Enhanced module functionality tests initially had a 78.2% pass rate, improving to 100% after resolving word-spotting and randomised body language issues. Middleware and application layers, using the same infrastructure, processed and visualised telerehabilitation data, showing 90.9 % functionality success. Adjustments to the nodes' input resolved concerns with enabling/disabling buttons, achieving 100 % functionality. Real-world experiments with CwA showed promising results, with the system transmitting 91.8% of MQTT packets accurately before resolving initial packet loss. The IoRT system’s CwA responses and eye gaze measurements matched the video analysis well, and the System Usability Scale (SUS) feedback rated the guardian and therapist dashboards’ usability as excellent, both above 80 SUS score. In conclusion, the proposed system proven to be robustly feasible, functional, and performs well in the set experiment environments. It successfully achieving targeted objectives and addressing the current system inefficiencies. This innovative system greatly benefits the autistic community by improving remote intervention, bridging distances between children and therapists, and setting a new standard for telerehabilitation.