In recent years, vehicular Ad Hoc networks(VANETs) has become a popular research area. Many attractive applications can be developed in VANETs to improve driving safety, avoid traffic jams, provide entertainment services for the passengers, etc. Data delivery is a critical issue in VANETs because most applications require efficient message exchanges between vehicles to perform well. However, to design an efficient and reliable data delivery protocol is a very challenged task since vehicles move fast and consequently network topology changes rapidly. Nowadays, more and more vehicles are equipped with GPS navigation systems which plan driving routes(trajectories) for drivers before their departure. In this paper, we utilize those trajectories to assist data delivery. When a source vehicle has data to be transmitted, other vehicles' trajectories are exploited to predict encounter probability between vehicles and use the probability to find a data delivery path. We classify the encounters between vehicles into six cases and analyse each encounter probability. We then present how to find a minimum hop path and a minimum delay path. Simulations are constructed to evaluate performance of the proposed scheme. Compared to existed data delivery, our scheme has lower delivery delay and higher delivery ratio, and the number of transmission hop count can be reduced.