Outage-Aware Online Prediction Control for Securing UAV-Aided Communication

Considering an unmanned aerial vehicle (UAV)-aided multi-user system with multiple mobile eavesdroppers, this paper aims to develop an outage-aware online control strategy to counter the imperfect information of mobile user equipments (UEs) and mobile eavesdroppers. To perform online control effectively, a novel online prediction model is proposed, which models the trajectory of UEs as a stochastic Gauss-Markov mobility model and the trajectory of eavesdroppers as a worst-case model based on the current imperfect location information. The online control problem is formulated as a new multi-stage optimization framework, in which we perform online control by solving per-slot prediction problems and a resource allocation problem for the initial point. In the per-slot prediction problem, we jointly optimize the predictive trajectory and transmit power for the upcoming time slot (TS) by maximizing the minimum achievable secrecy rate (ASR) among all UEs. For resource allocation at the initial TS, we optimize transmit power by maximizing the minimum ASR among all UEs. The per-slot prediction problem and the resourceallocation problem are all non-convex problems hence they are intractable in terms of computational complexity. To address them, we propose an efficient inner-approximation methods, by applying successive convex approximation (SCA) methodand S-procedure as well as Bernstein-type inequality. Numerical results demonstrate the superiority of our proposed online control scheme over the conventional benchmark schemes which include existing offline schemes and online scheme without outage based prediction as well as shortest-path navigation scheme.