Evolution strategies for reinforcement learning - OpenAI Roboschool