Improving the Expected Performance of Self-Organization in a Collective Adaptive System of Drones using Stochastic Multiplayer Games

Abstract

The Internet-of-Things (IoT) domain will be one of the most important domains of research in the coming decades. Paradigms continue to emerge that can employ self-organization to capitalize on the sheer number and variety of devices in the market. In this paper, we combine the use of stochastic multiplayer games (SMGs) and negotiation within two collective adaptive systems of drones tasked with locating and surveilling intelligence caches. We assess the use of an ordinary least squares (OLS) regression model that is trained on the SMG's output. The SMG is augmented to incorporate the OLS model to evaluate integration configurations during negotiation. The augmented SMG is compared to the base SMG where drones always integrate. Our results show that the incorporation of the OLS model improves the expected performance of the drones while significantly reducing the number of failed surveillance tasks which result in the loss of drones.