Network congestion analysis of gravity generated models

The network topology has lately proved to be critical to the appearance of traffic congestion, with scale-free networks being the less affected at high volumes of traffic. Here, the congestion dynamics are investigated for a class of networks that has experienced a resurgence of interest, the networks based on the gravity model. In addition, supplementary to the standard paradigm of uniform traffic volumes between randomly interacting node pairs, more realistic gravity traffic patterns are used to simulate the flows in the network. Results indicate that depending on the traffic pattern, the networks have different tolerance to congestion. Experiment simulation shows that the topologies created on the basis of the gravity model suffer less from congestion than the random, the scale-free or the Jackson–Rogers ones under both random and gravity traffic patterns. The congestion level is found to be approximately correlated with the network clustering coefficient in the case of random traffic, whereas in the case of gravity traffic such a correlation is not a trivial one. Other basic network properties such as the average shortest path and the diameter are seen to correlate fairly well with the congestion level. Further investigation on the adjustment of the gravity model parameters indicates particular sensitivity to network congestion. This work may have practical implications for designing traffic networks with both reasonable budget and good performance.