A Sequential Game Approach for Computation-Offloading in an UAV Network

Small drones are currently emerging as versatile nascent technology that can be used in exploration and surveillance missions. However, most of the underlying applications require very often complex and time-consuming calculations. Although, the limited resources available onboard the small drones, their mobility, the computation delays and energy consumption make the operation of these applications very challenging. Nevertheless, computation-offloading solutions provide feasible resolves to mitigate the issues facing these constrained devices. In this context, we address in this paper the problem of offloading highly intensive computation tasks, performed by a fleet of small drones, in order to improve the energy overhead and decrease the execution delay. We adopt a theoretical methodology based on a sequential game where three different types of players (drone, base station and edge server) carry out the heavy computation tasks. Compared to literature, as far as we know, we are the first to consider a computation-offloading problem with three different devices. Each player has a set of possible strategies, depending on the previous actions that the other players might undertake in a sequential game. Furthermore, we prove the existence of a Nash Equilibrium and design an offloading algorithm that converges to this optimal point. Extensive simulations gave promising results where the sequential game based model outperforms comparable approaches in terms of global utility, which pledges the best possible tradeoff between energy consumption and achievable delay.