Improved deep reinforcement learning for robotics through distribution-based experience retention | IEEE Conference Publication | IEEE Xplore