Trajectory generation for robotic pick-and-place task using nested double-memory deep deterministic policy gradient.