Self Learning Agent-1

Mangalmay Institute of Engineering and Technology
Greater Noida
Presentation On
Self Learning Game using
Deep Reinforcement Learning
Batch Members: Guided by :

Saisha Gulia-(2007860100068) Ms. Srinkhala Maddhesia
Samarth Dixit-(2007861520016) (Assistant Professor)

Shubham Saxena (20078611520017)
Introduction
Self-learning games harness the power of
Deep Reinforcement Learning to create
autonomous
game-playing agents.
Through DRL, these agents from experience,
adapt strategies, and achieve high
performance without human intervention.
These AI systems continuously improve,
showcasing the potential of artificial
intelligence in gaming and beyond.
Problem Statement
Develop an autonomous self-learning game agent using
Deep Reinforcement Learning (DRL) to achieve high
performance in a dynamic gaming environment. The
objective is to design an agent that can learn to play and
excel at a specific game without explicit human
intervention or supervision.
Literature Review
Advancements in self-learning games using DRL
• Recent advances in self-learning games using DRL, citing Mnih et al.
(2015), Silver et al.(2016), and Vinyals et al.(2019).
• Review highlights key contributions, methods, and outcomes, showing

DRL algorithm evolution and game challenges tackled.
• Future focus: boosting self-learning game agent abilities, broadening DRL

use in varied gaming contexts.
Objectives
•Autonomous Learning
•High Performance
•Adaptability
•Generalization
•Exploration & Exploitation
•Efficiency & Speed
•Robustness
•Scalability
Proposed Methodology
1. Environment Design and Integration
2. Deep Reinforcement Learning Architecture
3. Training Pipeline
4. Reward Based learning
5. Evaluation and Validation
6. Fine-tuning of learning agent

Results
• Training Performance: The plot displays the agent's scores
across 500 training episodes, initially low as the agent learns.
• Agent's Behavior: During testing, the agent applies learned
strategies to play the game more efficiently.
• Conclusion: Utilizing deep reinforcement learning, we trained an
AI agent for a paddle game, mastering paddle control and ball
bouncing over time. This project illustrates reinforcement
learning's applicability for training AI in interactive tasks and
games, with effective visualization aiding in result
communication.
IMPLEMENTATION RESULT
This Paddle game made
using DRL Shows a clear
progression from poor initial
performance to mastery of
the game. The agent learns
to optimize its actions
through continuous
interaction with the
environment, guided by the
reward structure, ultimately
demonstrating effective
decision-making skills in the
paddle game context.
Conclusion
Self-learning games utillizing
Deep Reinforcement Learning
demonstrate remarkable
potential in creating
autonomous game-playing
agents. Further research and
development in this field
promise to revolutionize
gaming experiences and AI
capabilities.
Future Work
• Multi-Agent Systems
• Transfer Learning and Generalization
• Curriculum Learning and Adaptive Difficulty
• Incorporating Human Feedback
• Robustness and Safety
• Interpretability and Explainability
• Real-world Applications
• Collaborative Research Initiatives

References
• 1. Haarmkinfotech, "Deep Reinforcement Learning: Training AI to Play Games and Solve Complex Problems,"
Medium.
• 2. Ujjwal Upadhyay, "Solving Unity Environment with Deep Reinforcement Learning," Towards Data Science.
• 3. V. Mnih et al., "Human-level control through deep reinforcement learning," Nature, vol. 518, no. 7540, pp.
529–533, Feb. 2015.
• 4. O. Vinyals et al., "Grandmaster level in StarCraft II using multi-agent reinforcement learning," Nature, vol.
575, no. 7782, pp. 350–354, Nov. 2019.
• 5. E2B, "The State of AI Agents: Reliability, SDKs, Benchmarking, and Market Trends."
• 6. GeeksforGeeks, "Agents in Artificial Intelligence."
• 7. Kaggle, "Intro to Game AI and Reinforcement Learning."
• 8. Silver, D., Singh, S., Precup, D., & Sutton, R. S., "Deep Reinforcement Learning in Large Discrete Action
Spaces," in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 4, pp. 3006-3013, 2020.
‫استعمال کریں‬
‫‪Thank You‬‬

Self Learning Agent-1

Uploaded by

Copyright:

Available Formats

Self Learning Agent-1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Self Learning Agent-1

Uploaded by

Copyright:

Available Formats

Mangalmay Institute of Engineering and Technology

Batch Members: Guided by :

Samarth Dixit-(2007861520016) (Assistant Professor)

• Review highlights key contributions, methods, and outcomes, showing

• Future focus: boosting self-learning game agent abilities, broadening DRL

•Exploration & Exploitation

•Efficiency & Speed

2. Deep Reinforcement Learning Architecture

4. Reward Based learning

5. Evaluation and Validation

6. Fine-tuning of learning agent

• Transfer Learning and Generalization

• Curriculum Learning and Adaptive Difficulty

• Incorporating Human Feedback

• Robustness and Safety

• Interpretability and Explainability

• Collaborative Research Initiatives

• 6. GeeksforGeeks, "Agents in Artificial Intelligence."

• 7. Kaggle, "Intro to Game AI and Reinforcement Learning."

You might also like