Abstract⌗

Wireless ad hoc networks rely on exchanging information among the nodes to find a feasible route between source and destination. Routing protocols require periodically flooding the entire system with the updated state of the network. Routing protocols define the rules and semantics of how this information is spread throughout the network and can be classified into proactive and reactive protocols, both of which eventually sends out control packets to the network. However, with the increase of the scale of the network, the amount of packets exchanged uses resources that could be otherwise used for application messages. In this paper, we explore the use of machine learning techniques to restrict the flooding radius of control packets. We define the problem objective as a regression problem and use neural networks to model it. The model can be used in networks that share similar attributes, suggesting that transfer learning can be used to re-purpose the initial model. The results show that using machine learning reduces the amount of resources used to maintain the network state up to date, which could be subsequently used to improve the overall performance of the networked system. We anticipate machine learning to be the starting point for more sophisticated models of cross-layer routing mechanisms.

Resources⌗

Bibtex⌗

@inproceedings{regis2021deep-learning,
    title={Deep-learning assisted Cross-Layer Routing in Multi-hop Wireless Network},
    author={Paulo Alexandre Regis, Suman Bhunia, Amar Nath Patra, and Shamik Sengupta},
    year={2021},
    publisher={IEEE},
    address={New Orleans, LA, USA},
    booktitle={None},
}