We propose a novel learning-based solution for motion planning of physically-based humanoid climbing that allows for fast and robust planning of complex climbing strategies and movements, including extreme movements such ...