Our Bootstrapping of Parameterized Skills Through Hybrid Optimization in Task and Policy Spaces, by Jeffrey Frederic Queißer, and Jochen J. Steil has just been published in Frontiers in Robotics and AI, section Humanoid Robotics. It presents a novel
approach for representing and incrementally optimizing a library of DMP skill by using CMA-ES in both the parameter
space of DMP and a high-level, but low-dimensional paramentrization of the skill with application to reaching to targets
through a grid. For your open access copy, go to paper link.