Using competition to control congestion in autonomous drone systems

Abstract

With the number and variety of commercial drones and UAVs (Unmanned Aerial Vehicles) set to escalate, there will be high future demands on popular regions of airspace and communication bandwidths. This raises safety concerns and hence heightens the need for a generic quantitative understanding of the real-time dynamics of multi-drone populations. Here, we explain how a simple system design built around system-level competition, as opposed to cooperation, can be used to control and ultimately reduce the fluctuations that ordinarily arise in such congestion situations, while simultaneously keeping the on-board processing requirements minimal. These benefits naturally arise from the collective competition to choose the less crowded option, using only previous outcomes and built-in algorithms. We provide explicit closed-form formulae that are applicable to any number of airborne drones N, and which show that the necessary on-board processing increases slower than N as N increases. This design therefore offers operational advantages over traditional cooperative schemes that require drone-to-drone communications that scale like N2, and also over optimization and control schemes that do not easily scale up to general N. In addition to populations of drones, the same mathematical analysis can be used to describe more complex individual drones that feature N adaptive sensor/actuator units.