Human-Robot Trust Assessment From Physical Apprehension Signals

To help enable robot-augmented production, where robots assist production staff to relieve them of repetitive and strenuous tasks, the purpose of the research is enabling real-time human-robot trust assessment by inferring decreases human trust from signs of physical apprehension. To ensure safe and productive human-robot collaboration we have to ensure an appropriate level of trust in the robot, as too much trust can lead to dangerous situations, whereas too little trust can lead to loss in productivity. The main hypothesis is that if the user experiences a decrease in trust, they will increase their distance from the robot by stepping or leaning away from it.

A series of experiments were performed using a Rethink Robotics Sawyer robot and an augmented reality enabled human-robot collaboration cell, using projection to display task critical information within the shared work space. Participants performed repeated tasks with the robot, midway through which the robot would disrupt the participants' expectations in order to decrease their trust towards it. Their movements were assessed using an infrared camera for body tracking to correlate it with decreases in trust.