Machine Learning of Semi-Autonomous Intelligent Mesh Networks Operation Expertise
- Alex Bordetsky, Naval Postgraduate School, CENETIX, Monterey, California, United States
- Carsten Glose, German Armed Forces (Bundeswehr), ESEP, Koblenz, Germany
- Steve Mullins, Naval Postgraduate School, CENETIX, Monterey, California, United States
- Eugene Bourakov, Naval Postgraduate School, CENETIX, Monterey, California, United States
AbstractOperating networks in very dynamic environments makes network management both complex and difficult. It remains an open question how mesh or hastily formed networks with many nodes could be managed efficiently. Considering the various constraints such as limited communication channels on network management in dynamic environments, the need for semi-autonomous or autonomous networks is evident. Exploitation of machine learning techniques could be a way to solve this network management challenge. However, the need for large training datasets and the infrequency of network management events make it uncertain whether this approach is effective for highly dynamic networks and networks operating in unfriendly conditions, such as tactical military networks. This paper examines the feasibility of this approach by analyzing a recorded dataset of a mesh network experiment in a highly dynamic, austere military environment and derives conclusions for the design of future mesh networks and their network management systems.
Return to previous page