earticle

영어

Simulation methodology is preferred approach to test vehicular networking protocols and to perform overhead analysis compared to high affordability and inconvenience of the real testbed. The confidence of simulation results is greatly dependent on the adoptive mobility model, i.e. whether or not the vehicular mobility modeling could reproduce the real vehicle moving pattern and could capture the intrinsic spatial as well as temporal relationships among vehicles. From the perspective of macro- and micro-scale, this paper proposes a series of statistical characteristics to assess whether the existing mobility models reflect a regular pattern. Micro-characteristics cover individual statistics, including driving duration time, driving distance, parking duration time, and link holding time. Simulations show that diverse mobility models share similar micro-characteristics in the complementary cumulative distribution function, but still display innegligible differences in distribution curves. Macro-characteristics refer to overall correlations, such as average velocity, total driving distance, total driving time, average neighbors' degree, and average accelerating time. Results show that the investigated mobility models behave with similar phase transition, but with different critical transition points. Marco-statistics follow Poisson distribution, but with completely different average values. The work is expected to help researchers better understand simulations and design context-aware vehicular algorithms.