Multi Agent Architecture for Automated Health Coaching

Area of Science:

• Health Informatics
• Software Engineering
• Human-Computer Interaction

Background:

• Effective health coaching software requires addressing privacy, modularity, scalability, individualization, data integration, transferability, coordination, and flexibility.
• Wearable sensors generate valuable data for personalized health interventions.

Purpose of the Study:

• To propose a novel, generic multi-agent architecture as a template for health coaching applications utilizing wearable sensors.
• To demonstrate the flexibility and adaptability of the proposed architecture in real-world health coaching scenarios.

Main Methods:

• Development of a generic multi-agent architecture with analyzer and communication modules.
• Implementation of diverse functionalities including goal formation, planning, scheduling, event detection, learning, and data collection.
• Creation of two distinct health coaching systems based on the proposed architecture: a static Fitbit-based system and a dynamic Apple Watch-based system.

Main Results:

• The proposed architecture successfully supports various functionalities essential for effective health coaching.
• A static health coaching system was developed using the Fitbit platform to promote physical activity at preset times.
• A dynamic health coaching system was developed using the Apple Watch platform, adapting interventions to encourage physical activity and reduce sedentary behavior.

Conclusions:

• The proposed multi-agent architecture provides a flexible and scalable template for developing diverse health coaching applications.
• The successful implementation of both static and dynamic systems validates the architecture's adaptability to different platforms and coaching strategies.
• This architecture facilitates the integration of wearable sensor data for personalized and effective health coaching.