Reinforcement-Learning Deep Recurrent Q learning Take successive image as Input of Network By Reccurent Network, Agent can predict Ball moving and move to the direction to block the ball Structure Image -> CNN -> RNN -> Q value