The Use of Reinforcement Learning in Gaming The Breakout Game Case Study.pdf

Abstract

This paper provides a comparative analysis between Deep Q Network (DQN) and Double Deep Q Network (DDQN) algorithms based on their hit rate, out of which DDQN proved to be better for Breakout game. DQN is chosen over Basic Q learning because it understands policy learning using its neural network which is good for complex environment and DDQN is chosen as it solves overestimation problem (agent always choses non-optimal action for any state just because it has maximum Q-value) occurring in basic Q-learning.

Basic things about reinforcement learning

“…5. Reward:If agent does an action close to the target, it will be rewarded and the opposite will be punished, and the reward can be received immediately or delayed [2].…”

“…Q-Learning is one algorithm of RL, the targer of Q-Learning is quite similar with RL, which is to find the optimal policy by using action-value function Q to approximate optimal action-value functioin Q [2]. Q-Learning will store the value of action-value functioin for combining each possible action with each state to form a table which called Q-table.…”

“…In RL, the destination of an agent is to get maximum total reward, therefore if we want to solve some realistic problem while the agent reach its target, we have to set a good reward function [2]. However, it can be knotty sometimes.…”