In VANETs, vehicle-to-vehicle communication is a challenge due to the fast topology changing, frequent disconnections, and highly variable traffic densities. This paper describes a Trajectory-based data Forwarding scheme with future Neighbor Prediction (TFNP) in autonomous driving vehicular environments. The most important characteristic of the environments is that the prediction for future position of vehicles could be sufficiently accurate. With the trajectory information of all autonomous vehicles, the future neighbors of each vehicle can be properly identified. The packet forwarding ability of the future neighbors is evaluated and forwarding sequences are thus constructed. When an autonomous vehicle needs to send data, it will transmit packets based on the constructed forwarding path. The simulation results demonstrate the effectiveness and efficiency of the developed scheme in the autonomous driving vehicular environments.