Text this: Improved double DQN with deep reinforcement learning for UAV indoor autonomous obstacle avoidance