Speeding Up Autonomous Learning by Using State-Independent Option Policies and Termination Improvement