AoI-minimal Clustering, Transmission and Trajectory Co-design for UAV-assisted WPCNs

This paper investigates the long-term average age of information (AoI)-minimal problem in an unmanned aerial vehicle (UAV)-assisted wireless-powered communication network (WPCN), which consists of a static hybrid access point (HAP), a mobile UAV, and many static sensor nodes (SNs) randomly distributed on multiple islands. The UAV first is fully charged by the HAP, and then flies to each island to charge SNs and receive data from them. Before running out the energy in battery, the UAV flies back to the HAP to offload the received data and be fully charged again. Due to the finite battery capacity of the UAV, it is impossible for the UAV to traverse all the islands to collect all the data from SNs for once flight. We are thus inspired to divide islands into multiple clusters so that the UAV could traverse all the islands in each cluster. The key factors affecting the long-term average AoI contain the hovering duration, the flying duration, and the amount of data from each island reflected by the number of SNs on each island. Therefore, we formulate the long-term average AoI-minimal problem by jointly optimizing the transmit power of SNs, clustering of islands, and UAV's flight trajectory, subject to the battery capacity of the UAV. Since the optimization problem is NP-hard, there are no standard methods to solve it optimally in general. To tackle this problem, we decouple it into two subproblems: the power allocation subproblem for SNs, and the joint clustering of islands and UAV's flight trajectory design subproblem, which is much more perplexed and complicated owing to the tight coupling between them. To solve the first subproblem, we propose a hybrid TDMA and NOMA (HTN) protocol that takes advantage of the two protocols. To solve the second subproblem, we propose a clustering-based dynamic adjustment of the shortest path (C-DASP) algorithm, which is composed of three sub-algorithms, i.e, a proposed merging-aided K-means clustering (MaKMC) algorithm, the particle swarm optimization (PSO) algorithm employed to find the shortest path in each cluster, and a proposed dynamic adjustment (DA) algorithm taking into account the number of SNs on each island. Simulations verify the effectiveness and superiority of the proposed HTN protocol and C-DASP algorithm.