Основні поняття
Bayes-POMCP optimizes human-robot team performance through adaptive interventions in mixed-initiative settings.
Анотація
The content discusses the development of a computational model, Bayes-POMCP, to enhance human-robot team performance in suboptimal scenarios. It explores the impact of different robot intervention styles on team performance and user preferences through two user studies. The results show that Bayes-POMCP outperforms heuristic policies and an adversarial baseline, improving both team performance and user satisfaction.
Abstract:
- Effective human-agent teaming requires robots to adapt to human abilities.
- Most prior works assume near-optimal teammates, but real-world collaboration involves suboptimal agents.
- The study develops a Bayesian approach for enhancing mixed-initiative collaborations between humans and robots.
- User studies demonstrate improved objective and subjective measures with the proposed approach.
Introduction:
- Human-agent teaming leverages unique capabilities of humans and AI agents.
- Real-world situations involve suboptimal performance due to uncertainty.
- Robots need a Theory of Mind to infer human teammates' mental states for effective collaboration.
- Mixed-initiative interactions are crucial for maximizing team performance.
Data Extraction:
- Our user studies show that user preferences and team performance vary with robot intervention styles.
- The proposed Bayes-POMCP approach enhances objective team performance (𝑝< .001) and subjective measures like trust (𝑝< .001) and likeability (𝑝< .001).
Статистика
Our user studies show that user preferences and team performance indeed vary with robot intervention styles.
The proposed approach enhances objective team performance (𝑝< .001) and subjective measures, such as user’s trust (𝑝< .001) and perceived likeability of the robot (𝑝< .001).
Цитати
"Robots need to develop a Theory of Mind to infer human teammates’ mental states."
"Our proposed approach maximizes human-robot team performance in real-time."