In order to evaluate vehicle collision avoidance systems, and to optimize the warning algorithm, this paper proposes a serial of indexes, including the effective warning rate and the wrong warning rate. According to the actual road conditions, simulation scenarios and experimental procedures are designed, for the experimental group and the control group, with and without warning system. A simulation evaluation method is used to obtain the corresponding indexes. According to the method, a platform for the warning algorithm simulation evaluation is built by the secondary development of the microscopic traffic simulation software VISSIM based on Visual Basic 2005. Various functions are implemented, including the parameter input, the result output at the VB port, the simulation process control, and the automatic statistical analysis. Honda and PATH algorithms are evaluated on the platform. It is shown that the latter algorithm has a higher effective warning rate and lower wrong warning rate as compared with the former one. The evaluation results are consistent with other related studies, which shows that the method is an effective quantitative analysis evaluation method with a low cost for warning algorithms.