Trust, but Verify: Autonomous Robot Trust Modeling in Human-Robot Collaboration

摘要

In this work, we propose a computational robot trust model based on the predictability of the human partner as the main factor that impacts robot trust. Using this model and based on the robot’s knowledge about the task, the robot switches its behavior between conservative and normal, and adjusts the implemented safety mechanism as a function of the current robot trust level. We illustrate the proposed model impact on the outcome of the collaboration using a simple scenario. The results show a significant improvement in safety conditions of the human in collaboration with a robot at the cost of justifiable team performance reduction.