Smart Routing in Urban Wireless Ad Hoc Networks Using Graph Attention Network-Based Decision Models
Abstract
Cellular mobile ad hoc networks represent a part of Wireless Ad hoc Network (WANET) that has received much attention in the recently developed landscapes of urban communication networks because of its real-time support for decentralized and scalable communication systems with mobile nodes. However, challenges become apparent, and they stem from the features of the urban environment where traffic changes dynamically and nodes are densely placed while random obstacles maybe in the way at any time. This work introduces an original smart routing framework in which GATs have been integrated in an attempt to improve the routing decision of WANETs in urban environments. Through the application of GAT-based decision model, the network is modeled as a dynamic graph in which nodes are mobile devices, and the edges represent wireless communication links with contextual features and topological learning patterns that allow the calculation of attention weights of neighboring nodes depending on the real-life context of communication. Due to this adaptive learning procedure, the orientation of these routes, less latencies for routing, and better packet delivery ratios are made possible even if the requirements involve intricate and fastmoving urban territories. The above method for designing a routing protocol is compared with the conventional routing techniques with the help of simulations, where the existing techniques are seen to be outperformed in terms of scalability, flexibility, and energy consumption. In summary, this article demonstrates that the attention-based deep learning schemes may be able to transform routing protocols in complex urban ad hoc networks into intelligent self-organizing systems of smart cities.