Network Economics-enabled Edge Computing in UAV-assisted Public Safety Systems

Abstract

In this paper, the joint problem of agents, e.g., police officers, firefighters, etc., to Unmanned Aerial Vehicles (UAVs) association and optimal partial task offloading is addressed based on the principles of reinforcement learning and contract theory, respectively, in public safety scenarios. A two-layers approach is followed. At the first layer, the agents act as learning automata in order to learn their most beneficial UAV selection to optimize their long-term reward in terms of processing their offloaded data, while respecting their delay constraints and tolerance stemming from their requested computing service and the public safety scenario that they serve. At the second layer, a contract-theoretic model is proposed to determine the agents’ optimal amount of offloaded data to the selected UAV and the UAV’s optimal portion of allocated computing capacity to each agent’s computing tasks, while considering the urgency of the agents’ requested service. A detailed set of numerical and comparative simulation results demonstrates the drawbacks and benefits of the proposed framework under rea-life public safety scenarios.