Pedestrian-to-vehicle communications plays an important role in preventing pedestrians from traffic accidents by informing pedestrian position to drivers of nearby vehicles. Global Navigation Satellite System (GNSS), the de facto positioning method, does not work well in urban canyons due to obstruction of roadside skyscrapers. To address this issue, we previously proposed a base method, leveraging vehicles, besides GNSS satellites, as anchors for computing pedestrian position, in which received signal strength of the line-of-sight (LOS) component of radio signals from vehicles is used to compute pedestrian-vehicle distance. The base method greatly reduces position error when there are a sufficient number of vehicles. However, its performance degrades much when few vehicles are available. To solve this problem, in this paper, we extend the base method from two aspects: (i) add roadside units (RSUs) as anchors, and (ii) estimate the direction of arrival of signals from vehicles and RSUs for positioning by applying the MUSIC algorithm. We evaluate the extended method by 3D ray tracing simulation. Extensive results confirm the effectiveness of the extended method. In a scenario where the number of vehicles is limited, the position error in the base method, 1.94 m, is reduced to 0.28 m in the extended method.