We propose a newly pedestrian model for evaluating vehicle-pedestrian accidents using STREET we have been developing. The pedestrian model can simulate pedestrian-to-vehicle interactions by making pedestrians and drivers behave in their surroundings based on their cognition and decision. It reproduces the behavior of pedestrians by setting up action rules governing combinations of positions of pedestrians, their modes of action, and cognition targets. Finaly, we have validated the pedestrian mode with an example of simulating basic operations regarding the ignoring of signals and vehicle-to-pedestrian right-turning gaps, thereby comfirming that the example exhibits a tendency that closely corresponds to the actual measurements.