User Profiling to Enhance Clinical Assessment and Human-Robot Interaction: A Feasibility Study

Area of Science:

• Robotics
• Human-Robot Interaction
• Gerontology

Background:

• Socially Assistive Robots (SARs) require personalized, human-like behavior for acceptance and trustworthiness.
• SARs can serve as companions, assistants, and valuable tools for caregivers, aiding in patient assessment and monitoring.

Purpose of the Study:

• To present and discuss a robot behavioral model using multimodal features for user profiling and interaction fine-tuning.
• To test the proposed model in a real-world environment with the SAR robot ASTRO, focusing on gait and physical interaction analysis.

Main Methods:

• Developed a robot behavioral model integrating sensing, perception, decision support, and interaction modules.
• Utilized multimodal features for both user profiling (decision support for caregivers) and social cue correlation (interaction fine-tuning).
• Tested the model with ASTRO, measuring body posture, gait cycle, and handgrip strength in ten older adults during a walking support task.

Main Results:

• The model accurately estimated gait parameters, handgrip strength, and torso angular excursion compared to standard instruments (p < 0.05).
• Unsupervised dimensionality reduction techniques (t-SNE, nMDS) effectively profiled users based on residual walking abilities.
• Sensory outputs were successfully combined in a perceptual model for user assessment.

Conclusions:

• The proposed robot behavioral model effectively profiles users and enhances human-robot interaction for Socially Assistive Robots.
• This approach offers a promising tool for caregivers, providing objective patient assessment and supporting personalized care.
• The study validates the use of SARs in real environments for monitoring and assisting older adults, improving their quality of life.