Uprising E-sports Industry: machine learning/AI improve in-game performance using deep reinforcement learning

With the quick development of machine learning, deep reinforcement learning will have a big influence on E-sports. It can be considered machine learning will help us train E-sports players easily and effectively and give coaches and players some new ideas to train and win the games. Flappy Bird is a game where the players try to keep the bird alive as long as possible and get a high score. Flappy Bird is a good example to prove that the thought is feasible. In this project, a flappy bird training AI is developed based on Q-learning and DQN. The game is played by the models obtained from deep reinforcement learning, and the game is also played by humans. Then get experimental data from these two ways and compare them. For the two ways of playing the game (by AI or manually), there are many similarities in the increased rate of scores as training sessions increase, which means AI can “teach” players how to train to get a higher score. It can be applied to skills and experience-based games and help us to train top players. Maybe it can also be applied to other fields, such as helping engineers escape from potential errors and accidents.