A DRL-assisted trusted geographic position-based routing protocol for FANETs with abnormal nodes.

Abstract

In this paper, we address the critical challenges of ensuring fast and reliable communications in flying ad hoc networks (FANETs) composed of uncrewed aerial vehicles. While promising for cooperative aerial missions, FANETs suffer from performance degradation and robustness issues in the presence of abnormal nodes. Furthermore, existing routing protocols often fail to achieve an effective balance between low delay and low energy consumption. To overcome these limitations, we propose a deep reinforcement learning-assisted trusted geographic position-based routing protocol (TGPRP). The proposed protocol formulates the routing decision process as a Markov decision process. Within this framework, node credibility quantifies the reliability of neighboring nodes. The state space is designed to integrate key features, including node position information and residual energy, while a feedback mechanism captures local network states. At the core of TGPRP is a deep value network, which is trained to select optimal next-hop nodes. This training is guided by a reward function that dynamically adapts to real-time network conditions. Extensive simulation study demonstrates that our TGPRP significantly outperforms existing routing schemes by reducing the average end-to-end delay, energy consumption and packet loss rate in scenarios with abnormal nodes. Moreover, TGPRP exhibits superior adaptability and robustness under varying network scales and different ratios of abnormal nodes. The protocol's capability for real-time perception and intelligent decision-making effectively ensures robust overall network performance in dynamic FANET environments.