Posts

Showing posts from 2021

Machine-Learning Agents: First Steps Towards Building a Model for Testing (Part 2)

Image
After some extensive testing for a period of four weeks, we found a lot of lot of issues with ML-Agents when it comes to game development in general. We have made a video on the topic after the end of the project, explaining our findings: -Amirali Mohayaee

Machine-Learning Agents: First Steps Towards Building a Model for Testing (Part 1)

Image
Initial plans have started for a project with my teammate Athanasios "Thanos" Athanasiadis on a project which uses Machine Learning using either heuristics or reinforced learning methods and implements an agent that would be able to react to player actions and playstyle. We hope that we can create a series of scenes to test a variety of agent models. Our rough timetable for the coming weeks will include research in conjunction with constant development of the scenes. These are not set in stone and can be subject to change depending on changes in scope and development on a weekly basis. (Alternative links to the spreadsheet here). Early production has begun on our project on Unity, using Unity's own Collaborate internal tools. While the toolset is relatively new to the team, it is similar to Git based options. We have begun creating some basic scenes to test behaviors. Implementation will begin once agents have been trained and brains developed. We will s