Learning from Demonstration for Goal-Driven Autonomy