A ROBotic INTelligent INTuitive and INTeractive platform for NAO-Mediated Autistic Healthcare

From manufacturing purposes to service tasks, the primary uses of robotics on a large scale currently involve highly specialized applications. These include performing extremely repetitive tasks in industrial settings, domestic floor cleaning, and robotic surgery. The success of robotics in these areas lies in the simplicity of use for the end user: industrial robots are pre-programmed, vacuum robots need only be directed to the area to be cleaned, and surgical robots are teleoperated by surgeons like advanced tools. Conversely, the main obstacle to the widespread use of robotic systems for non-repetitive tasks and operations not directly controlled by a human operator, as in surgery, is the difficulty faced by operators in defining the tasks the robots must perform. This is due to the lack of specific programming skills required for their use. This project aims to remove this barrier by developing a novel architecture that facilitates human-robot interaction. Users will command complex robot behaviors using natural language descriptions or practical demonstrations. Large Language Models (LLMs) serve as the core technology, translating human requests into robot-interpretable commands. LLMs, a type of artificial intelligence, leverage machine learning to understand and generate human language and have already demonstrated impressive capabilities in generating code from textual descriptions. The project's primary application focuses on controlling social robots, specifically the NAO robot. This small humanoid robot is used in healthcare settings to interact with patients with autism spectrum disorders (ASD), facilitating diagnosis and treatment. Pilot studies conducted at the PASCIA center of the Modena Polyclinic during the past year have shown promising results, indicating potential reductions in cardiology visit duration and improved feasibility. However, a large-scale statistical analysis is still lacking, and the aim is not only to understand the overall effects of using a social robot on the success of visits but also to evaluate the impact of specific actions performed by the robot, which, thanks to the proposed architecture, the doctor can define independently and very quickly. Additionally, the project aims to assess how the proposed control architecture can enable personalized interaction between the patient/doctor mediated by the robot, extending the use of this device to other medical fields, such as dentistry.