Conflicts in human operator - unmanned vehicles interactions

Abstract

In the context of the supervision of one or several unmanned vehicles by a human operator, the definition and the dynamics of the shared authority among these agents is a major challenge. Indeed, lessons learned from modern aviation reveals that authority sharing issues between aircrews and on-board processes are remarkable precursors of air accidents (twenty accidents in the last twenty years). The analysis of these events highlights that the authority of the on-board processes is designed a priori and fails to adapt in case of conflict with the aircrew's actions. Moreover the poor design of the HMIs (e.g. : there is no dialogue between artificial and human agents) and the complexity of the interactions may lead the aircrews to lose situation awareness and to enter a perseveration syndrome. We present the basic concepts of an approach aiming at dynamically adjusting the autonomy of an agent in a mission relatively to its operator, based on formal detection of conflict. An experimental set-up is under construction to assess our hypotheses.