Local Roots A Tree Based Subgoal Discovery Method to Accelerate Reinforcement Learning

Subgoal discovery in reinforcement learning is an effective way of partitioning a problem domain with large state space. Recent research mainly focuses on automatic identification of such subgoals during learning, making use of state transition information gathered during exploration. Mostly based on the options framework, an identified subgoal leads the learning agent to an intermediate region which is known to be useful on the way to goal. In this paper, we propose a novel automatic subgoal discovery method which is based on analysis of predicted shortcut history segments derived from experience, which are then used to generate useful options to speed up learning. Compared to similar existing methods, it performs significantly better in terms of time complexity and usefulness of the subgoals identified, without sacrificing solution quality. The effectiveness of the method is empirically shown via experimentation on various benchmark problems compared to well known subgoal identification methods