Empirical and simulation study on traffic oscillation characteristic using floating car data

Traffic oscillation is a very important phenomenon in traffic flow studies and its characteristics are usually used to validate the developed traffic flow models and theories. In this paper, we firstly conduct an empirical study based on floating car data, focusing on the characteristic of traffic oscillations in terms of the overall amplitude and the spectrum of speed/acceleration as well as the relation between the characteristic quantities. It is found that the amplitude of both speed oscillations and acceleration oscillations grows with the increase of traffic speed. The spectrum analysis shows that the most energy would concentrate on a narrower band with the increase of standard deviation of speed/acceleration, and the periodicity in the speed oscillation is more salient than that in the corresponding acceleration oscillation. Then, three typical car-following models are tested: one is the deterministic model, while the other two are the stochastic models with different mechanisms of stochasticity. The simulation results indicate that the two stochastic models outperform the deterministic model in reproducing the observed oscillation characteristics qualitatively and quantitatively.