Artificial Intelligence Dominates Starcraft 2 with Reinforcement Learning

Table of Contents

1. Introduction
2. Understanding Starcraft 2
3. The Objective of Starcraft 2
4. Building an Army of Attack Units
5. Collecting Resources
6. Base Building
7. Strategizing and Knowledge
8. Tuning Muscle Memory
9. Using Libraries for Automation
10. Machine Learning and Starcraft 2
11. Importing Reinforcement Learning
12. Determining Inputs and Outputs
13. The Importance of Inputs in Neural Networks
14. Visuals as Inputs
15. Other Graphical Representations as Inputs
16. Using Vectors as Inputs
17. Outputs in Deep Reinforcement Learning
18. Actions and Decisions
19. Example Actions in Starcraft 2
20. Implementing Actions in the Model
21. Creating a Mini-Map as Model Input
22. Drawing the Minerals
23. Marking Enemy Locations
24. Marking Own Nexus and Structures
25. Marking Vespene Gas
26. Drawing Units
27. Coloring the Void Ray Unit
28. Finalizing the Mini-Map
29. Testing the Model with Random Actions
30. The Role of Rewards in Agent Learning
31. Challenges of Reward Mechanisms
32. Positive and Negative Rewards
33. Rewarding Attack Actions
34. Incentivizing Elimination of the Enemy
35. Limitations and Potential Improvements
36. Connecting Stable Baselines 3 and Starcraft 2
37. Converting to OpenAI Gym Environment
38. Handling Communication Between Systems
39. Training the Model with Stable Baselines 3
40. Tracking Overall Rewards and Wins
41. Evaluating Reward Mechanisms for Success
42. Exploring the Winning Rate
43. Continuing the Journey of Reinforcement Learning in Starcraft 2
44. Expanding to Micro Tasks
45. Conclusion and Future Possibilities

Introduction

Starcraft 2 is a popular multiplayer game that requires strategic thinking and quick decision-making. In this article, we will explore the world of Starcraft 2 and how to use deep reinforcement learning to improve gameplay. We will Delve into the gameplay mechanics, the objective of the game, and the strategies involved. Additionally, we will discuss the process of building an army, collecting resources, and establishing a base. Furthermore, we will explore the concept of muscle memory and the use of libraries to automate gameplay. We will then delve into the world of machine learning and its application in Starcraft 2. Lastly, we will cover the implementation of reinforcement learning and the creation of a mini-map as model input. So let's dive in and discover how we can use deep reinforcement learning to conquer the world of Starcraft 2.