Steady-State of The SLAW Mobility Model

Abstract

Movement patterns of mobile nodes significantly affect the performance of a Mobile Ad hoc Network (MANET) routing protocol. Thus, it is essential that the mobility traces generated by a mobility model closely match the trace data collected from experiments. In addition, the probability distribution of mobile nodes (over the simulation area) varies with time before it comes to a steady-state. In other words, distribution of mobile nodes in steady-state is independent of their initial position. If traces generated by a mobility model are used before its steady-state is reached, this variability in distribution may lead to misleading results, called initialization bias. Thus, for credible MANET simulations, it is required that a mobility model used (1) is realistic (i.e., its mobility traces closely match the experimental data) and (2) starts in a steady- state (i.e., there is no initialization bias). In this work, we analyze a recently published realistic mobility model, called SLAW. The SLAW mobility model is based on real GPS traces collected from five outdoor sites. Previous work with SLAW is done by discarding the first few hours of simulation time; however, discarding initial data does not guarantee that the model has reached its steady-state. Thus, the main contribution of our work is that we provide methods for sampling from the steady- state distributions of mobile nodes' locations and pause- times. Sampling from the steady-state distributions allows the SLAW mobility model to start in a steady-state and thus, avoids initialization bias related to the mobility model.