Predicting Future Agent Motions for Dynamic Environments