Title: Cluster head selection and routing using fractional deep deterministic policy gradient in FANET

Authors: Kathuroju Srinivas; Malaiyappan Nandhini

Addresses: Department of Computer Science, Pondicherry University, Puducherry, 605014, India ' Department of Computer Science, Pondicherry University, Puducherry, 605014, India

Abstract: Deep learning (DL) offers significant advancements in the selection and routing of cluster heads (CHs), which are characterised by their dynamic environments and high mobility. By utilising supervised learning, drones can be classified based on battery life, communication range, and historical performance. Here, a fractional deep deterministic policy gradient (FDDPG) enabled multi-agent actor-critic network is devised for CH selection and routing in flying ad hoc networks (FANETs). At first, a FANET simulation is conducted. Then, flying node metrics computation for every node is conducted based on multi-objective parameters. Multi-agent actor-critic Network is employed to select CH and to determine the routing protocol, where the training is conducted using FDDPG approach, which is the integration of fraction calculus (FC) and deep deterministic policy gradient (DDPG). It is recognised that the developed model attained energy consumption of 55.596 J, a distance of 73.293 m, a delay of 0.489 ms, and a trust of 0.879.

Keywords: FANET; flying ad hoc network; UAVs; unmanned aerial vehicles; cluster head selection; routing; DDPG; deep deterministic policy gradient.

DOI: 10.1504/IJAACS.2025.150826

International Journal of Autonomous and Adaptive Communications Systems, 2025 Vol.18 No.6, pp.553 - 579

Received: 21 Feb 2025
Accepted: 14 Jun 2025
Published online: 23 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article