Time-Oriented Joint Clustering and UAV Trajectory Planning in UAV-Assisted WSNs: Leveraging Parallel Transmission and Variable Velocity Scheme

Unmanned aerial vehicles (UAVs) have been regarded as an efficient approach for collecting data in wireless sensor networks (WSNs), benefited from their mobility and flexibility. In this work, we investigate the data collection problem in UAV-assisted WSNs. In order to improve data collection efficiency, we first propose a multi-scenario parallel data collection scheme which allows data packets being transmitted through various modes/links simultaneously. Then, addressing the importance of completing data collection within a short time duration, we formulate a constrained optimization problem which minimizes the data collection time of the sensor nodes (SNs) by jointly designing UAV flight trajectory, cluster head mode selection, SN clustering strategy and UAV velocity. To resolve the optimization problem, we first consider the data transmission performance between SNs and present an SN clustering scheme based on a modified K-means algorithm. Given the clustering strategy, the optimization problem is then converted into three sub-problems, i.e., CH mode selection, UAV trajectory design, and flight velocity optimization. Firstly, jointly considering the data collection time of the cluster heads in various transmission modes and the spectrum resources of the sink node, we propose a greedy method-based CH mode selection scheme. Then, we map the UAV trajectory optimization problem as a traveling salesman problem and propose a simulated annealing-based algorithm to determine the flight trajectory for the UAV. Finally, by applying discrete time segment scheme, the UAV velocity optimization subproblem is transformed into a sequence of convex flight time minimization problems and a segment optimization-based flight velocity control strategy is presented. Numerical results reveal that the proposed data collection algorithm can achieve
$25{\mathrm{\% }}$
and
$12{\mathrm{\% }}$
performance gains comparing to the existing algorithms and the benchmark scheme, respectively.