Towards A Machine Learning-Based Framework For Automated Design of Networking Protocols
Hannaneh Barahouei Pasandi · 2019
Today, due to the increasing demands on wireless communications and new emerged technologies, new protocols will be designed faster than before. By evolving network technologies as well as increasing demands of modern applications, general-purpose protocol stacks are not always adequate and need to be replaced by application tailored protocols. To cope with the emergence of various device characteristics and application requirements, complex and custom design of high performance networking protocols is needed. The current methods for protocol design are mainly human-based and thus are burdened with various limitations. Design of new protocols is time-consuming and requires a specialized knowledge that is not trivial to acquire. This is especially limiting in the context of modern networking domain, i.e., IEEE 802.11 protocol that is continuously evolving nowadays to meet new requirements and conditions through the addition of new amendments. Furthermore, once a protocol is designed, it lacks optimal adaptability and flexibility to changes in the environment, since contemporary communication scenarios display dynamic and non-stationary properties. Changes in network are so fast and frequent that no human-based mechanisms can follow them accurately. Finally, current approaches are limited to human perception and understanding of this field, thus limiting the potential for extracting new and unexpected insights during the protocol design process. Therefore, replacing this inefficient human-based protocol designing process by a novel paradigm that enables rapid design of efficient, flexible, and high performance protocols that intelligently adapt to different device characteristics, application requirements, user objectives, and network conditions is highly desired.