Combining Multi-Objective Planning With Reinforcement Learning To Solve Complex Tasks In Environments With Sparse Rewards