Zanzibar Bazaar - A Multi-Agent System Project

This repository contains our implementation of the Zanzibar Bazaar multi-agent system using the JADE framework. In this scenario, multiple merchant agents compete in a spice market, while a BazaarAgent (Game Master) orchestrates rounds, announces events, and calculates player scores.

---

Table of Contents

1. Overview
2. Project Structure
3. Requirements
4. Running the Project
5. Agents Description
6. Author and Contact

---

Overview

Zanzibar Bazaar simulates a historical spice market where merchants attempt to maximize profits by trading, forming alliances, and reacting to random events (storms, port taxes, new trade routes). The system demonstrates:

• BDI-inspired agents (AdvancedPlayerAgent) that track beliefs (prices, events), form desires (maximize profit, sabotage rivals), and execute intentions (propose trades, accept or reject offers).
• A BazaarAgent that manages each round, updates prices based on supply/demand, and introduces game-ending conditions.
• A simpler agent (SimplePlayerAgent), which uses minimal logic (random or coin-flip acceptance), serving as a baseline for comparison against the BDI-capable PlayerAgents.

---

Project Structure

A typical directory layout might look like this:

zanzibar-bazaar/
├─ docs/
|   ├─ Project_Presentation.pdf // slides
|   └─ Project_Report.pdf // final report
├─ simulation/
|   ├─ output.txt // output log
|   ├─ screen_record.mkv // video capture of the running simulation
|   ├─ sniffer_snapshot_file.txt // canva of message exchange, can be opened from the sniffer
|   └─ sniffer_message_list // human readable message exchange list
├─ src/
│   ├─ zanzibar/bazaar/
│   │   ├─ BazaarAgent.java // The Game Master agent
│   │   ├─ AdvancedPlayerAgent.java // BDI-like merchant agent
│   │   └─ SimplePlayerAgent.java // Simple merchant agent with minimal logic
└─ README.md // This file

• BazaarAgent.java: Coordinates rounds, broadcasts events, calculates scores, and applies taxes or price changes.
• AdvancedPlayerAgent.java: A BDI-like agent with inventory tracking, event forecasting, alliances/rivalries, and negotiation.
• SimplePlayerAgent.java: A simpler agent that randomly makes or accepts trade proposals, illustrating minimal strategy.

---

Requirements

• Java 8+ (tested up to Java 11 or higher)
• JADE 4.6.0 or compatible version
  Download from: JADE website

To compile and run, ensure you have both jade.jar and jadeTools.jar (Sniffer agent) in your classpath.

---

Running the Project

1. Clone or Download the Repository

git clone https://github.com/YourUsername/zanzibar-bazaar.git
cd zanzibar-bazaar

2. Compile the JADE Agents

You can use a build tool (Maven/Gradle) or directly compile with javac, ensuring the classpath references JADE:

javac -cp .:path/to/jade.jar zanzibar/bazaar/*.java

3. Launch the JADE Runtime

Run JADE's main container, specifying the classpath. For example:

java -cp .:path/to/jade.jar jade.Boot -gui

Alternatively, you can start the system via a script or an IDE run configuration.

4. Start the Agents

You can either start them from the JADE RMA (Remote Monitoring Agent) or specify them at launch. For example:

java -cp .:path/to/jade.jar:./build zanzibar.bazaar.BazaarAgent -maxRounds 5
java -cp .:path/to/jade.jar:./build zanzibar.bazaar.AdvancedPlayerAgent
java -cp .:path/to/jade.jar:./build zanzibar.bazaar.SimplePlayerAgent

Adjust the classpath and package names as needed.

5. Monitor with Sniffer (Optional)

Add -services jade.core.messaging.TraceService or use the JADE GUI’s Sniffer to watch message exchanges.

---

Agents Description

1. BazaarAgent

• Registration: Registers in the Directory Facilitator as bazaar-master.
• Responsibilities:
  • Conducts the marketplace rounds:
    • Update Player List (DF check).
    • Announce Round Start.
    • Announce Prices (and random events).
    • Request Sales.
    • Collect Sales and update the scoreboard.
    • End Round and adjust market prices.
  • Terminates the game with a game-over message after reaching maxRounds.

---

2. PlayerAgent (BDI-like)

• Beliefs: Maintains knowledge of current prices, events, inventory, and known players.
• Desires: Aims to maximize profit, form alliances, and sabotage rivals.
• Intentions:
  • NegotiationInitiatorBehavior to propose trades.
  • NegotiationResponder to accept/reject trades.
  • AllianceProposer and AllianceResponder.
  • SabotageBehavior (optional).
• Decision-making: Uses naive forecasting to decide whether to hold or sell each spice.

---

3. MerchantAgent (Simple)

• Registration: Similar DF registration as zanzibar-player.
• Behavior:
  • Maintains a budget and a small inventory.
  • Randomly proposes or accepts trades (50% acceptance).
  • Picks random quantities to sell when responding to the sell-request.

These three agent types interact in the same environment, showcasing how different negotiation and forecasting strategies yield diverse outcomes.

---

Author and Contact

• Tommaso Tragno - tommaso.tragno@gmail.com
• Duarte Alexandre Pedro Gonçalves - duarte.dapg@gmail.com

For any questions or suggestions, feel free to open an issue or reach out by email. Pull requests to improve or extend the simulation are also welcome!