In modern days, UAV (Unmanned Aerial Vehicle) are being extensively used in various fields like military, healtcare, security, government sectors, supervision, home delivery agents, etc. They significantly enhance the potential of IoT devices by processing their data. However, issues like efficient trajectory planning, security and privacy protection, task scheduling, a delegation of tasks from one UAV to another in multi-UAV systems, etc., require rigorous research and analysis. In this paper, we consider a multi-UAV multi-IoT network, which is divided into certain hexagonal cells. Each cell consists of some IoT devices and a UAV to process the data those IoT devices collect from the environment. IoT devices in each cell are grouped into clusters so that UAVs hover only above cluster heads and midpoints of cell boundaries to collect and delegate tasks whenever required. Provision of both single and multi-hop task delegation exists in our system. The schedule is formed based on the UAV's next intended arrival time as computed by cluster heads in the cell using the Auto-Regressive Moving Average (ARMA) model. Timestamps of visiting midpoints of cell boundaries are fitted between predicted next arrival times of visiting cluster heads. Energy consumption and time of transition from one hovering point to another can be optimized. However, the most energy-efficient solution may not necessarily be the most delay efficient. A multi-objective optimization technique is applied to identify the Pareto-optimal front and select the best possible solution. Simulation results show that compared to other trajectory planning algorithms viz. Dijkstra's and HEA, our proposed technique saves much more energy (approx 78% over the other two) and time (42% over the other two).
