In recent years, many packet forwarding protocols adopting the social concept have been proposed for vehicular social networks (VSN). However, these protocols do not consider that vehicle trajectory information and the social relationship between vehicles at different periods may be distinct significantly. In this paper, we propose the Socially-Aware Trajectory-based Routing (SATR) protocol, which analyzes vehicle trajectories and social relationships at different time intervals, designs the vehicle encounter graph and constructs the corresponding packet relay path graph to improve the performance of packet forwarding in VSN. We execute simulations for well-known trajectory-based and social-based routing protocols by the Opportunistic Network Environment (ONE) simulator under different parameter values. These results exhibit that SATR significantly improves low accuracies of social-based protocols to choose relay nodes and reduces high computation complexities of trajectory-based ones.