Digital Twin AI Fitness Coach: An Intelligent Multi-Agent System for Personalized Exercise Guidance
Vahdati, Monica HamlAbadi, Kamran Gholizadeh Laamarti, Fedwa Kumar, Devansh El Saddik, Abdulmotaleb
Vahdati, Monica
HamlAbadi, Kamran Gholizadeh
Laamarti, Fedwa
Kumar, Devansh
El Saddik, Abdulmotaleb
Supervisor
Department
Computer Vision
Embargo End Date
Type
Conference proceeding
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
This paper presents the Digital Twin AI Fitness Coaching System (DTAIFC), an innovative platform that integrates a multi-agent architecture with advanced pose estimation and Large Language Models (LLMs) for intelligent fitness coaching. The system combines OpenPose-based skeletal tracking with CrewAI-orchestrated agents to deliver real-time feedback across text, voice, and image modalities. A specialized pose comparison algorithm evaluates exercise form and provides personalized guidance. DTAIFC leverages LangSmith for performance tracking, enabling analysis of pose accuracy, safety, and feedback quality. Our implementation demonstrates effective coordination among agents, each contributing to a seamless coaching experience. Experimental results validate the system's ability to deliver context-aware, personalized feedback through both synchronous and asynchronous interactions. The evaluation, based on pose estimation metrics and multi-modal interaction assessment, confirms DTAIFC's effectiveness in providing accurate, adaptive coaching, marking a significant advancement in AI-driven fitness instruction.
Citation
M. Vahdati et al., “Digital Twin AI Fitness Coach: An Intelligent Multi-Agent System for Personalized Exercise Guidance,” Proceedings of the 1st International Workshop on Multi-Sensorial Media and Applications, pp. 18–26, Oct. 2025, doi: 10.1145/3728485.3759171.
Source
MSMA '25: Proceedings of the 1st International Workshop on Multi-Sensorial Media and Applications
Conference
The 33rd ACM International Conference on Multimedia
Keywords
Digital Twin, Multi-agent System, Agentic AI, Multimodal Interaction, Large Language Model, AI Fitness
Subjects
Source
The 33rd ACM International Conference on Multimedia
Publisher
Association for Computing Machinery