Application
Autonomical control of the formation of WMNs to enable node density control, frequency diversity and cooperative, multi-domain software-defined networking (SDN) on large-scale wireless mesh networks.
Key Benefits
- Enhanced capacity of wireless mesh networks.
- Robust to changes in nodes (new nodes going up, old nodes going down, etc.).
- Enables multi-domain centralized network management strategies such as software-defined networking.
Market Summary
A wireless mesh network (WMN) is a communications network composed of nodes that provide access to the network (e.g. routers) organized in a mesh of linkages. Mesh nodes cooperate, forwarding data on behalf of each other to achieve a wide coverage despite using short-range wireless links. As opposed to a traditional wired network, in a WMN, only one node needs to be physically wired to a network connection. WMNs show great promise for enhancing connectivity over large regions at low cost, since fewer wires are required, but are hampered by poor capacity scaling and hardened network control mechanisms. In networking, capacity is defined as the amount of traffic that a network can handle at any given time and in WMN, increasing the number of nodes does not increase its capacity. Therefore, for WMNs to be widely adopted at large-scales, there is a need to first improve capacity scaling. To facilitate its application in diverse settings, it requires more flexible and programmable network control, the mechanism deciding how a given data traffic must traverse the WMN.
Technical Summary
Emory researchers developed a set of self-organizing, self-healing, and self-optimizing autonomic agents, the SMART class of agents, that takes an existing WMN node placement and evolves it into interconnected partitions with improved capacity scaling. The autonomic agent in this model operates in a distributed manner to generate a partition set with limited intra-partition latency and controlled node density given the limited network partition diameter and the connectivity degree of the nodes. SMART agents also perform self-configuration, self-organization, self-healing, and self-optimization independently and using local information (autonomic computing principles). In multiple simulations ran each with thousands of nodes, SMART showed fast convergence to stable configurations of partitions and that the agent design is robust to connectivity failures among the nodes and other changes in node placement or number. The design does not rely on any centralized network formation mechanism, turning appropriate for application in networks at large-scales.
Developmental Stage
The SMART class of agents has been successfully tested in a simulated environment with one thousand nodes, using multiple topologies advent from realistic urban settings (Microsoft Buildings dataset).
Publication: Gramacho, S. et al. (2019). 2019 International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), 175–82.
Read our featured innovation.
