Research And Application Of Game Artificial Intelligence System Based On Machine Learning Methods

In the field of machine learning,how to let agents learn directly from high-dimensional sensory inputs like speech and vision is one of the long-standing challenges of reinforcement learning(RL)domains.Before Deep Q-Learning Network(DQN)put forward,Many successful RL applications based on hand-crafted features combined with linear value functions or policy representations.Obviously,the performance of those systems gravely rely on the quality of the feature representation.A agent can learn what to do to maximize a numerical reward signal from interaction with the environment with a DQN and can successful learn control policies with a steady manner such as some Atari 2600 games.Convolution networks and Q-Learning,DQN proposes a new idea in Reinforcement learning in game Artificial Intelligence domain.However reinforcement learning presents several challenges from a deep learning perspective.Deep Q-Network's performance falls far below human level in situations that exist delayed rewards and require planning under uncertainty within long-time horizon to optimize strategies.This paper use Deep Recurrent Q-Network(DRQN),a combination of a Long Short Term Memory(LSTM)and a Deep Q-Network with a optimized Asynchronous methods for training and buiding a system to verify the effectiveness of the model.Experimental results show that the agent system we build has significantly improved In 3D FPS game(Doom).