Distributed Reinforcement Learning with Neural Networks for Robotics