How Does Trust in Simulations of Drone Failures Compare with Reality?

Abstract

As drone capabilities expand, they are being used in far more safety-critical applications. Drones, like any other electronic device, are prone to malfunctions, and these malfunctions which can result in collisions usually occur from errors in the underlying algorithms or from noisy communications. With the use of these drones in closer proximity to humans than ever before, commonly referred to as social robot navigation, these drones need to operate while not harming any person. In addition, to allow wide adoption in safety critical applications in light of possible malfunctions, an investigation of humans' trust in drones, has emerged as a topic of increasing importance in recent years. Understanding how our trust can change, while using a drone, can be useful in determining how to minimize this loss in trust. Unfortunately, testing humans' trust in drones is expensive and risky in the situations where humans are using drones in an enclosed space. Therefore, it becomes necessary to model flight failures via simulations. This paper works to validate if simulated trust produces equivalent results compared to real-life test results.