lnu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
D3QN-Powered Adaptive Coaching System for Game Performance Enhancement
Karnavati University, India.
Karnavati University, India.
Karnavati University, India.
Linnaeus University, Faculty of Technology, Department of computer science and media technology. Linnaeus University, Sweden.ORCID iD: 0000-0002-2487-0866
2025 (English)In: 2025 International Conference on Electrical Communication and Computing Technologies Iconecct 2025, Institute of Electrical and Electronics Engineers (IEEE) , 2025Conference paper, Published paper (Refereed)
Abstract [en]

The esports industry's explosive expansion has resulted in compelling interest in high quality and scalable coaching solutions for players. However, most automated coaching solutions employ behavioral cloning, thereby binding players to mimic what an expert does including non-optimal behaviors rather than seeking victory-oriented strategies. The current project attempts to address this severe limitation by creating a novel, end-to-end AI coaching system based on offline Deep Reinforcement Learning (DRL) that provides genuine strategy optimization. The architecture of the system consists of three distinct and multi-functional layers of integration. First, a rigorous Perception Pipeline will ingest raw, high frequency logs of gameplay data and create a feature-rich standardized experience replay buffer establishing the necessary data foundation. Second, the Analysis Layer will use a state-of-the-art Dueling Double Deep Q-Network (DDQN) - Dueling decouples state-value through action-advantage analysis while Double DQN mitigates overestimation bias and stabilizes training. Third, the innovative Recommendation Layer incorporates a Large Language Model (LLM) as a communicative interface reshaping the numerical and qualitative Q-value assessment of the DRL agent into relevant and actionable coaching advice for the player. The experimental results confirmed the performance of the system being used. The DRL model successfully showed stable convergence during training, as can be seen by the decrease in Huber loss and the steady increase in average maximum Q -values indicating that winning strategies that earned good rewards were employed. This project demonstrated conclusively that the combination of the analytical performance of DRL and the natural language processing capabilities of LLMs is a compelling and viable approach to developing esports players in a way that is increasingly sophisticated and scalable.

Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers (IEEE) , 2025.
Keywords [en]
AI Coaching, Artificial Intelligence, Behavioural Cloning, Computer Vision, Esports, Gameplay Analysis, Large Language Models, Replay Parsers, Supervised Learning
Identifiers
URN: urn:nbn:se:lnu:diva-150914DOI: 10.1109/iCONECCT67014.2025.11470047Scopus ID: 2-s2.0-105037992956ISBN: 9798331574154 (print)OAI: oai:DiVA.org:lnu-150914DiVA, id: diva2:2115101
Conference
2025 International Conference on Electrical Communication and Computing Technologies Iconecct, Gwalior, India, December 26-28, 2025
Available from: 2026-11-04 Created: 2026-11-04 Last updated: 2026-11-04

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Ghayvat, Hemant

Search in DiVA

By author/editor
Ghayvat, Hemant
By organisation
Department of computer science and media technology

Search outside of DiVA

GoogleGoogle Scholar

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf