| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.92 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A inteligência artificial em videojogos é uma área de investigação de longa data. É um conceito importante em muitos jogos e estuda como utilizar tecnologias de IA para alcançar o desempenho a nível humano durante o jogo. No entanto, quando se trata de IA e videojogos, a Reinforcement Learning tem de ser mencionada. RL define os agentes que enfrentam os problemas que aprendem a tomar boas decisões apenas através da acção e observação.
Este projecto centra-se na integração de um algoritmo de Machine Learning chamado Reinforcement Learning no desenvolvimento de um videojogo do género Tower Defense. O projeto foi desenvolvido pelo motor Unity3D que incorpora um agente que utiliza a técnica RL para simular o comportamento de um jogador humano e continuar a melhorá-lo, com base em experiências de jogo anteriores, até ser totalmente optimizado com uma pontuação imbatível pelo jogador médio. O agente irá imitar o comportamento de um humano, comprando, actualizando e colocando torres enquanto obtém a pontuação mais alta, utilizando o menor número de moedas. Além disso, o relatório irá também rever vários conceitos de Aprendizagem Automática, incluindo o Processo de Decisão de Markov e o Q-Learning.
Artificial intelligence in video games is a longstanding research area. It is a major concept in a lot of games and it studies how to use AI technologies to achieve human-level performance when playing games. However, when it comes to AI and video games, Reinforcement Learning has to be mentioned. RL defines the problem-facing agents that learn to make good decisions through action and observation alone. This project focuses on integrating a Machine Learning algorithm called Reinforcement Learning in the development of a video game of the Tower Defense genre developed by the Unity3D engine that incorporates an agent that uses the RL technique to simulate the behavior of a human player and keep on improving it, based on previous game experiences, until it’s fully optimized with a score unbeatable by the average player. The agent will imitate the behavior of a human, buying, upgrading, and placing towers while getting the highest score by using the lowest number of currencies. Moreover, the report will also review several Machine Learning concepts, including Markov-Decision Process and Q-Learning.
Artificial intelligence in video games is a longstanding research area. It is a major concept in a lot of games and it studies how to use AI technologies to achieve human-level performance when playing games. However, when it comes to AI and video games, Reinforcement Learning has to be mentioned. RL defines the problem-facing agents that learn to make good decisions through action and observation alone. This project focuses on integrating a Machine Learning algorithm called Reinforcement Learning in the development of a video game of the Tower Defense genre developed by the Unity3D engine that incorporates an agent that uses the RL technique to simulate the behavior of a human player and keep on improving it, based on previous game experiences, until it’s fully optimized with a score unbeatable by the average player. The agent will imitate the behavior of a human, buying, upgrading, and placing towers while getting the highest score by using the lowest number of currencies. Moreover, the report will also review several Machine Learning concepts, including Markov-Decision Process and Q-Learning.
Description
Mestrado em IPB-ESTG
Keywords
Defesa da torre Inteligência artificial Reinforcement learning Machine learning Redes neuronais