Event-triggered robot self-assessment to aid in autonomy adjustment

Area of Science:

• Robotics
• Artificial Intelligence
• Human-Robot Interaction

Background:

• Human-robot teams face complex tasks requiring adaptable autonomy.
• Dynamic environments and robot limitations necessitate dynamic adjustment of autonomy levels (LOAs).
• Assessing robot competency is key to determining appropriate LOAs.

Purpose of the Study:

• To develop a framework for dynamically adjusting robot autonomy based on competency.
• To introduce a Model Quality Assessment (MQA) metric to detect unexpected robot observations.
• To present an Event-Triggered Generalized Outcome Assessment (ET-GOA) algorithm for competency-based autonomy adjustments.

Main Methods:

• Proposed a Model Quality Assessment (MQA) metric comparing robot observations to model predictions.
• Developed an Event-Triggered Generalized Outcome Assessment (ET-GOA) algorithm using MQA changes to assess competency.
• Validated MQA and ET-GOA in simulated and live robot navigation tasks.

Main Results:

• MQA effectively detected unexpected observations from the robot.
• ET-GOA algorithm's computational cost and accuracy were analyzed across various thresholds and perturbations.
• Experimental results demonstrated the algorithm's responsiveness to changes in robot performance.

Conclusions:

• The ET-GOA algorithm facilitates informed decisions for adjusting robot autonomy.
• Assessing robot competency through MQA is a viable approach for managing dynamic autonomy.
• This research supports more effective human-robot teaming in complex operational scenarios.