HospitalMAS — Multi-Agent Hospital Simulation

HospitalMAS is an academic project developed for the Agents and Multi-Agent Systems course, as part of the Master's Degree in Artificial Intelligence at the University of Minho.

The project explores the use of Multi-Agent Systems (MAS) to model, coordinate and simulate hospital workflows in a distributed and autonomous way.

The repository contains both stages of the work:

• TrabalhoInvestigacao (Research Work): state-of-the-art analysis of Multi-Agent Systems in healthcare;
• TrabalhoPratico (Practical Work): implementation of a multi-hospital simulation using autonomous agents.

Final Grades

Component	Description	Grade
Research Work	Multi-Agent Systems in Healthcare	17/20
Practical Work	Multi-Agent Hospital Simulation	17/20

Project Motivation

Modern healthcare systems are complex, dynamic and resource-constrained environments. Hospitals must coordinate doctors, nurses, patients, consultation rooms, emergency services, exams, surgeries and inpatient beds, often under time-critical conditions.

Traditional centralized systems may struggle to adapt to unexpected events, such as sudden emergency demand, resource unavailability or changes in clinical priority.

This project investigates how Multi-Agent Systems can support more flexible, decentralized and adaptive hospital management.

General Objective

The main objective of HospitalMAS is to design and implement a simulation where hospital entities are represented as autonomous agents capable of:

• communicating with each other;
• negotiating resource allocation;
• managing patient flows;
• prioritizing urgent cases;
• scheduling routine consultations;
• coordinating exams, surgeries and hospitalization;
• monitoring hospital load through supervisors and a dashboard.

Research and Practical Development

The project was developed in two stages.

Research Work

The research phase presents a state-of-the-art review of Multi-Agent Systems applied to healthcare.

It studies several application domains, including:

• clinical triage;
• hospital resource management;
• patient monitoring;
• medical IoT;
• clinical decision support;
• digital twins;
• LLM-based healthcare agents;
• agent communication protocols.

The research also proposes a conceptual architecture for a hospital MAS, which later served as the basis for the practical implementation.

Practical Work

The practical phase implements a working simulation of a distributed hospital environment.

The implemented system includes:

• two hospitals;
• a unified central triage;
• routine consultations;
• emergency consultations;
• exams and medical tests;
• surgeries;
• hospitalization;
• medical staff agents;
• room and equipment agents;
• coordinator agents;
• supervisor agents;
• real-time dashboard.

Main Technologies

The project uses:

• Python;
• SPADE;
• XMPP;
• FIPA-inspired Contract Net protocol;
• FastAPI;
• Uvicorn;
• HTML/CSS/JavaScript dashboard;
• asynchronous agent communication.

Repository Structure

.
├── TrabalhoInvestigacao/
│   ├── README.md
│   ├── TrabalhoInvestigacaoG5.pdf
│   ├── ApresentacaoTI.pdf
│   └── MIA ASMa_2526_Enunciado_TI.pdf
│
├── TrabalhoPratico/
│   ├── README.md
│   ├── main_sim.py
│   ├── dashboard.py
│   ├── requirements.txt
│   ├── FLUXOS_AGENTES.md
│   ├── src/
│   ├── static/
│   ├── data/
│   ├── diagrams/
│   ├── docs/
│   └── tests/
│
└── README.md

System Overview

The practical system follows a distributed multi-agent architecture.

Patients
  ↓
Central Triage Agent
  ↓
Hospital Supervisor Agents
  ↓
Coordinator Agents
  ↓
Resource Agents
  ↓
Medical Acts and Hospital Flow

Each hospital contains specialized coordinator agents responsible for different workflows. These coordinators interact with resource agents to allocate doctors, nurses, rooms, equipment, operating rooms and hospitalization beds.

Main Agent Types

The system includes several types of agents:

• Patient Agents: represent patients entering the hospital system;
• Central Triage Agent: selects the hospital with the lowest relevant load;
• Consultation Coordinator Agents: manage routine consultation scheduling;
• Emergency Coordinator Agents: manage urgent patients and priority-based care;
• Exam Coordinator Agents: coordinate medical exams and diagnostic tests;
• Surgery Coordinator Agents: allocate surgeons and operating rooms;
• Hospitalization Coordinator Agents: manage rooms, beds and nursing resources;
• Doctor Agents: represent medical professionals;
• Nurse Agents: represent nursing staff;
• Room and Equipment Agents: represent physical hospital resources;
• Supervisor Agents: monitor hospital state and provide data to the dashboard.

Coordination Strategy

The system uses a hybrid coordination approach.

Routine consultations are managed through a centralized future-slot scheduling mechanism, ensuring realistic appointment planning according to doctor shifts, room availability and administrative working hours.

Emergency care, exams, surgeries, triage and hospitalization use a negotiation strategy inspired by the FIPA Contract Net protocol, where coordinators request proposals from available resources and select the best candidate according to availability, suitability and load.

Dashboard

The project includes a web dashboard that allows the user to observe the current state of the simulation.

The dashboard displays:

• hospital resources;
• active queues;
• routine and emergency patients;
• recent events;
• hospital load;
• resource occupation;
• system logs.

Academic Context

Developed at:

University of Minho Master's Degree in Artificial Intelligence Agents and Multi-Agent Systems Academic Year 2025/2026

Disclaimer

This project is an academic simulation and is not intended for real clinical use. It does not replace certified hospital management systems, medical decision-support tools or healthcare professionals.

License

This repository is intended for academic and educational purposes.