Towards a Learning-Based Framework for Self-Driving Design of Networking Protocols

Abstract

Networking protocols are designed through long-standing and hard-working human efforts. Machine Learning (ML)-based solutions for communication protocol design have been developed to avoid manual effort to adjust individual protocol parameters. While other proposed ML-based methods focus mainly on tuning individual protocol parameters (e.g. contention window adjustment), our main contribution is to propose a new Deep Reinforcement Learning (DRL) framework to systematically design and evaluate networking protocols. We decouple the protocol into a set of parametric modules, each representing the main protocol functionality that is used as a DRL input to better understand and systematically analyze the optimization of generated protocols. As a case study, we introduce and evaluate DeepMAC a framework in which the MAC protocol is decoupled into a set of blocks across popular 802.11 WLANs (e.g. 802.11 a/b/g/n/ac). We are interested to see which blocks are selected by DeepMAC across different networking scenarios and whether DeepMAC is capable of adapting to network dynamics.INDEX TERMS Communication protocols, deep learning, machine-generated algorithm, protocol design, reinforcement learning.This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2021.3061729, IEEE Access H. B. Pasandi, T. Nadeem: Towards a Learning-Based Framework for Self-Driving Design of Networking Protocols II. BACKGROUND HANNANEH BARAHOUEI PASANDI (Student Member, IEEE) received the master's degree from Université Grenoble Alpes. She is currently pursuing the Ph.D. degree with the Department of Computer Science, Virginia Commonwealth University. Her research interests include wireless communication, with an emphasis on the design and implementation of data-driven MAC protocols that not only consider conventional factors but also extract and include important environmental factors in designing protocols for IEEE WLANs.

Latte

“…The proxy server pre-fetches video segments of available video rates from the cloud for streaming to the receivers. The [3] that takes into account not only the conventional factors that impact MU-MIMO grouping such as SNR, number of spatial streams but also external factors such as number device characteristics and user mobility (the motion of individual users can be tracked by reading their explicit Channel State Information (CSI)). By considering these factors as the input to the RL agent, LATTE learns how to optimize the MU-MIMO grouping and mode selection for each environment such that the downlink throughput is maximized.…”

The Emergence of Wireless MAC Protocols with Multi-Agent Reinforcement Learning

“…In [2], a framework to design a protocol is proposed by considering the different functions a MAC protocol must perform. An RL agent designs a protocol by selecting which building function to use according to the network conditions.…”

“…This heterogeneity of wireless networks may represent a challenge to protocol design. Therefore, protocols tailored to specific applications may perform better than general-purpose solutions [2].…”

Wi-Fi Meets ML: A Survey on Improving IEEE 802.11 Performance with Machine Learning

“…An open challenge and a disruptive future approach would be to re-design these functionalities by explicitly embedding ML capabilities in them. Heuristic algorithms or hard-coded rules could be replaced by ML agents able to self-configure based on gathered experience [80], [321]. For example, in spatial reuse, the transmission power is adjusted following a set of predefined rules and this may unnecessarily limit the achievable throughput in some scenarios.…”