Towards more effective skill discovery in reinforcement learning by incorporating state reachability. — SciRadar