An ad hoc mobility model based on realistic human interactions

Abstract

Simulation is important for the validation of mobile ad hoc network protocols, and the effectiveness of simulation relies largely on what mobility model is used and how realistic the model is. Although there is an increasing amount of real traces in the public domain like CRAWDAD, the availability of them is still so limited that synthetic models are necessary for simulations. In order to simplify real movement patterns, most of the existing synthetic models generate movements with great randomness, and thus the initiative of people is not sufficiently reflected. Besides, instead of moving randomly, city residents or workers tend to regard only one location as more important than any other locations in many cases. For an individual, such a place may be his home in the neighborhood or his own desk in the office, and we call it a most important place (MIP), which is distinguished from other places. According to these real scenarios, we propose a new mobility model based on human interactions, taking into consideration human relationships, the distances of moving and the effect of MIPs. In the model, we quantify these factors in matrices and integrate them to calculate the probabilities of individuals selecting every possible destination. With the transfer-probability matrix determined, movements of each individual are equal to a Markov process, in which one location is viewed as one state of the Markov chain. Then we implement our model and give a dynamic demonstration of the moving nodes. To evaluate the reliability of our model, we use a CRAWDAD real trace as the baseline for comparison, and the result shows that our trace is close to reality.