From Service Dogs to Robots: A Bio-Inspired Framework for Intuitive Intent Prediction in Mixed Reality

Drawing on thousands of years of interspecies co-evolution, this study investigates the migration of natural human-service dog interaction paradigms to human-quadruped robot control (managing four-legged robotic systems). This study developed a multimodal intent prediction model using a Bayesian approach, trained on interaction data captured within a custom-built Mixed Reality (MR) environment. The model was integrated into an MR-based robot control system and evaluated through comparative user experiments. Results indicate that the proposed system significantly outperforms traditional speech-only and standard multimodal interfaces in operational efficiency, user emotional experience, and usability. This research provides a scalable framework for enhancing human-robot synergy in complex service scenarios, such as multi-robot coordination and emotional companionship, while establishing the MR interface as a versatile hub for intuitive robot control.