Shared autonomy for cognitive load regulation in autonomous wheelchairs

Shared autonomy for cognitive load regulation in autonomous wheelchairs

연구 분야: Artificial Intelligence

학회: Cognition, Technology & Work

This study advances human-machine partnership design in assistive mobility. It introduces a cognitive digital twin framework for autonomous wheelchair systems. We investigate dynamic collaboration mechanisms between users with motor impairments and autonomous driving systems, focusing on situational awareness regulation and adaptive autonomy allocation. Through Unity-based digital-twin simulation experiments, we evaluated multi-level human-machine interface (HMI) designs and interaction strategies, ranging from manual (L0) to conditional automation (L3) with 52 participants. Progressive information presentation, tailored to user cognitive pace and augmented by real-time interruption prompts in Advanced Driving Assistance Systems (ADAS) interfaces, significantly reduced reaction times by 23.5% and improved task accuracy for up to 98.1%. Shared and supervisory control strategies, with higher automation levels, markedly enhanced user experience and operational efficiency while lowering cognitive load. Physiological measures indicated minimal stress variation across strategies, suggesting cognitive load are primarily sourced from inherent information processing demands. These findings highlight the working mechanism of dynamic HMI layering and shared control in raising user-centric human-system partnerships, while proposing novel metrics for human-AI transparency in autonomous wheelchairs. This paves the way for safer, more efficient, and user-centered autonomous wheelchairs, promote the application of low-speed cooperative driving vehicles in public indoor scenarios such as hospitals, airports, and museums.