The relations of "go and stop" wave to car accidents in a cellular automaton with velocity-dependent randomization

In this paper we numerically study the probability P-ac of the occurrence of traffic accidents in the Nagel-Schreckenberg (NS) model with velocity-dependent randomization (VDR). Numerical results show that there is a critical density over which car accidents occur, but below which no car accidents happen. Different from the accident probability in the NS model, the accident probability in the VDR model monotonously decreases with increase of car density above the critical density. The value of the accident probability is only determined by the stochastic noise and the number of cars on road. In the stochastic VDR model with the speed limit v(max) = 1, no critical density exists and car accidents happen in the whole density region. The braking probabilities of standing cars and moving cars have different influences on the accident probability. A mean-field theory reveals that the accident probability is proportional to the mean density of "go and stop" wave per time step. Theoretical analyses give excellent agreement with numerical results in the VDR model. (C) 2007 Elsevier B.V. All rights reserved.