Context-Sensitive Proactive Robot: Empirical Insights and Computational Adaptation

A robot's ability to initiate help in a non-intrusive yet effective manner has become a critical requirement for trust and rapport in human-robot relationships. This thesis aims to understand and provide design implications for proactive robot behavior design through two stages: an empirical study to understand people’s perceptions, and the implementation of a context-aware computational framework. In the first stage, an empirical study (N=12) was conducted to investigate how the environment, proactivity types, and proactivity levels influence users' perception of robot proactivity. Through thematic qualitative and quantitative analysis, the study derives key empirical insights into user acceptance.In the second stage, this work proposes PROACH, a context-aware adaptation framework for mobile robots. PROACH adapts the robot's approaching speed and speech level of detail based on contextual factors including user engagement, cognitive load, and task properties. A within-subjects ablation study (N=12) was conducted to evaluate the framework against two baseline versions and an ablated version across various robot delivery types. The evaluation demonstrates that the joint optimization of speech and movement significantly enhances the robot's social appropriateness and perceived animacy, making the interaction appear more attentive and lively compared to unimodal adaptive or non-adaptive systems.