Programming Human-Drone Interactions: Lessons from the Drone Arena Challenge

Abstract

We report on the lessons we learned on programming human-drone interactions during a three-day challenge where five teams of drone novices each programmed a nanodrone to be piloted through an obstacle course using bodily movement. Center to the participants' learning process was the eventual shift from the deceptively simple idea of seamless human-drone interactions, to the reality of drones as non-predictable systems prone to crashes. This happened as participants had to first realize, then to deal with the limitations of the drone's resource-constrained hardware. Coping with these limitations was crucially complicated by the lack of appropriate programming abstractions, which led participants to focus on plenty of low-level, sometimes immaterial details, while losing focus on the ultimate objectives. We find concrete evidence of these observations in how participants handled the visibility problem in debugging drone behaviors, applied different defensive coding techniques, and altered their piloting practice. Our insights may inform further research efforts in drone programming, especially in the vastly uncharted territory of human-drone interactions.