Publication: Generating game character behaviors and animations through reinforcement learning
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering::Electrical and electronic engineering | |
| dc.contributor.author | Wan Siti Aaishah binti Wan Mohd Zaki | |
| dc.date.accessioned | 2025-06-05T08:38:31Z | |
| dc.date.available | 2025-06-05T08:38:31Z | |
| dc.date.issued | 2024-08 | |
| dc.description.abstract | Artificial intelligence (AI) has emerged as a major driving force in various industries and especially common in game animations. Usually, game characters were animated fully by animators and manually formatted into the game characters by programmers where the one animation will transition to another upon certain conditions. This have proven to be too laborious and limits the overall ability of a game character to react to random encounters. Hence, this thesis has proposed the use of artificial intelligence, which is deep reinforcement learning, a subset of machine learning (ML), to develop a self-learning non-player character (NPC) which can learn to react to randomized environment on its own. This project also experimented the different properties of neural network hyperparameters and their effects on the learning curve of physically different machine learning agents. Benefiting on open-source Unity ML-agent Toolkit, Python API and Anaconda console, this project is executed by first developing a Hummingbird agent to simply train it to perform basic maneuverer and collect nectars from flowers. Then, the Hummingbird agent along with a Ragdoll Walker agent was trained on different values of hyperparameters to evaluate and compare their learning performance. From this project, the Hummingbird agent has successfully learned on its own to perform desired actions. The effect of hyperparameter tuning towards the training performance on different agents has also been confirmed to not be necessarily proportional to the value of the hyperparameters. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/22063 | |
| dc.language.iso | en | |
| dc.title | Generating game character behaviors and animations through reinforcement learning | |
| dc.type | Resource Types::text::report::technical report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |