Formalizing Human-Robot Mutual Adaptation: A Bounded Memory Model.

Area of Science:

• Robotics
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Effective team collaboration is essential for complex tasks.
• Human-robot teams require mutual adaptation for optimal performance.
• Existing methods for human-robot adaptation have limitations.

Purpose of the Study:

• To propose a novel formalism for human-robot mutual adaptation in collaborative tasks.
• To introduce the bounded-memory adaptation model (BAM) for capturing human adaptive behaviors.
• To enable robots to adapt to human partners, improving team effectiveness and trust.

Main Methods:

• Developed the bounded-memory adaptation model (BAM) based on bounded memory assumptions.
• Integrated BAM into a partially observable stochastic model for robot adaptation.
• Conducted human subject experiments to evaluate the proposed formalism.

Main Results:

• The proposed formalism significantly improved the effectiveness of human-robot teams.
• Robot adaptation strategies were shown to be effective whether the human was adaptive or not.
• Human subject ratings for robot performance and trust were comparable to state-of-the-art cross-training methods.

Conclusions:

• The bounded-memory adaptation model (BAM) provides a robust framework for human-robot mutual adaptation.
• The formalism enhances human-robot team performance and maintains human trust.
• This approach offers a promising alternative to traditional human-robot team training practices.