A Proposal to Integrate Deep Q-Learning with Automated Planning to Improve the Performance of a Planning-Based Agent