Analysis of least-time and minimum-hop routing for clustered temporal networks

Networks of low-altitude multiple satellites (LAMS) have been envisioned to provide such diverse tasks as global positioning, paging, weather tracking, surveillance, and environmental observation and disaster assessment. One theoretical model for such dynamically reconfiguring networks, allowing the investigation of packet routing, is the clustered temporal network (CTN). The store-and-forward routing in such a network requires a quite different algorithmic approach to that of static networks which have been studied extensively. While solutions for optimal paths can be found only by NP complexity algorithms, a polynomial algorithm using spanning graphs in conjunction with the CTN model, and yielding very good, but not necessarily optimal, solutions have been previously found. The time-complexities of the algorithm using a full spanning graph, which yields optimal solutions relative to the CTN model, and a reduced spanning graph that yields good but sub-optimal solutions are investigated empirically. The algorithm using the reduced spanning graph is shown, through stochastic simulation and descriptive statistics, to be significantly faster than the algorithm using the full spanning graph. It is also shown that the space-complexity of the reduced spanning graph data structure is significantly less than the corresponding full spanning graph. Finally, the average path latencies and hop counts of the generated paths are compared for the full and reduced spanning graphs, and show very little difference. The studies are performed for networks of 5-80 nodes (satellites).