5 CONCLUSIONS
We have presented a maneuver predictor DBMP for highway driving scenarios. The algorithm is based on the Dynamic Bayesian Network, which combines structural prediction and dynamic features. The method has been validated using open dataset I-80. The results demonstrate that the proposed DBMP can effectively predict lane keeping and lane change over a long-term horizon. Moreover, the prediction criteria will greatly increase using proposed DBMP than naive BN. For the future work, we will investigate the motion prediction in other complex environments (e.g., urban intersections). The problem of unstable predictions will be solved in future studies.
