Control and prediction in hierarchical wireless networks

Abstract

Directional wireless networks (DWNs) using free space optical (FSO) and RF transmissions provide wireless backbone support for mobile communications in dynamic environments. The heterogeneous and dynamic nature of such networks challenges their robustness and requires self-organization mechanisms to assure end-to-end broadband connectivity. We have developed a framework to provide prediction and control strategies for assured network operation. We draw an analogy between a set of interconnected communication nodes and a molecule in which the bonds between atoms are representative of the links in the equivalent network. The dynamics of the network, and its optimization, can be analyzed by the methods of molecular dynamics. Links are modeled as bonds described by potential energy functions, such as the Morse potential, and a global description of the stability of the network can be obtained by a normal mode analysis (NMA). Effective “forces” act on nodes, which include the effects of power control, link length, and channel characteristics. A molecular re-arrangement or fragmentation occurs because this reduces the potential energy. In the same way a network can undergo topological reconfiguration, and an adaptive control strategy can be used to release, retain or reconfigure communication links for network performance optimization. Simulation results show the effectiveness of our self-organized control mechanism, where the physical topology reorganizes to maximize the number of source to destination communicating pairs. NMA of a network suffering degradation shows a correlation between anomalous eigenvalue behavior of the Hessian matrix describing the network and the improvement of network performance that can be achieved by topology change.