MACTS

Multi-Agent Control of Traffic Signals is a legacy MSE project that explores adaptive traffic-signal control with a small multi-agent system built around SUMO, RabbitMQ, and Python 2.7.

Project status: archived. This repository is preserved for reference and historical documentation. It is not under active development and should not be expected to run unchanged on a modern environment.

Looking for the maintained successor? See macts-next for the leaner, modernized follow-on repository.

Legacy project site: https://k0emt.github.io/macts/

Overview

MACTS models two coordinated traffic-light junctions in SUMO:

• JunctionSS (St Saviours)
• JunctionRKLN (Rose Kiln Lane)

The runtime design is message-driven:

1. A CommunicationsAgent launches SUMO, advances the simulation, reads detector data through TraCI, and publishes metrics and sensor events.
2. Two reactive planning agents (JSS_Reactive_Agent.py and JRKL_Reactive_Agent.py) consume junction sensor data and choose the next candidate phase.
3. A safety agent per junction validates signal progression and minimum timing before issuing traffic-light commands.
4. A MetricsAgent aggregates emissions, fuel, noise, halting vehicles, and mean-speed data, then persists a simulation summary to MongoDB.

The repository also includes design documents, project artifacts, UML diagrams, and earlier spike/prototype work from the original graduate project.

Repository layout

Path	Purpose
source/macts/	Primary legacy implementation: agents, SUMO configs, detector/network files, and RabbitMQ helper scripts
source/sumo_networks/	Additional SUMO network experiments and earlier simulation setups
source/spikes/	Exploratory prototype code and experiments
docs/	Published project artifacts and the legacy project website (docs/index.html)
portfolio/	Editable project documents, presentations, plans, and supporting coursework artifacts
uml/	UML models, diagrams, and sequence/activity illustrations
references/	Papers, reference material, and archived phase deliverables

Key implementation files

Within source/macts/:

• CommunicationsAgent.py — orchestrates SUMO, TraCI, RabbitMQ messaging, detector reads, and metric publication
• JSS_Reactive_Agent.py — reactive planner for St Saviours
• JRKL_Reactive_Agent.py — reactive planner for Rose Kiln Lane
• JSS_SafetyAgent.py / JRKL_SafetyAgent.py — safety validation and traffic-light command publication
• TrafficLightSignal.py — signal-state and signal-phase transition rules
• MetricsAgent.py — aggregates run metrics and stores them in MongoDB
• Core.py — shared agent, exchange, metric, and sensor-state primitives

Legacy technology assumptions

The codebase reflects its original 2011-2012 environment:

• Python 2.7
• SUMO with TraCI
• RabbitMQ (the bundled batch scripts reference RabbitMQ 2.8.1 on Windows)
• MongoDB
• Python packages such as pika and pymongo
• A largely Windows-oriented setup (*.bat helpers and hard-coded paths)

Important caveats:

• Several scripts use Python 2 syntax and are not Python 3 compatible as written.
• Some paths are hard-coded for the original workstation layout, such as C:\sumo-0.14.0\bin\ and C:\macts\source\macts\.
• Parts of the architecture are exploratory or incomplete; for example, collaboration/planning abstractions exist, but the implemented behavior centers on the reactive agents plus safety checks.

Running the legacy simulation

This project is best treated as a reconstruction target rather than a turnkey application. If you want to run it, start from source/macts/ and expect to adapt paths, package versions, and broker/database setup.

A likely legacy workflow is:

1. Install a Python 2.7 environment with pika, pymongo, SUMO, and TraCI bindings.

2. Start RabbitMQ and MongoDB.

3. Update the Windows helper scripts if your SUMO or RabbitMQ installation paths differ:

   • start-command-line.bat
   • rabbitmq_init.bat
   • rabbitmq_clean.bat

4. Create the RabbitMQ virtual host and users with rabbitmq_init.bat.

5. Declare exchanges with:

   python rabbitmq_create_exchanges.py

6. Start the long-running agents in separate terminals:

   python MetricsAgent.py
   python JSS_Reactive_Agent.py
   python JRKL_Reactive_Agent.py

7. Launch the simulation controller:

   python CommunicationsAgent.py low 300

The CommunicationsAgent.py usage pattern is:

python CommunicationsAgent.py [low|medium|full] MAX_NUMBER_SIMULATION_STEPS

Where:

• low, medium, and full select the SUMO route/configuration variant
• the final integer sets the maximum number of simulation iterations

Tests

The repository includes unit tests for signal progression logic in:

• source/macts/TrafficLightSignalTests.py

Most of the system behavior, however, depends on external legacy services and a matching SUMO environment.

Documentation and historical artifacts

The most useful archival materials are:

• docs/index.html — legacy project homepage
• docs/nehl_implementation_user_manual.pdf — archived user manual
• docs/nehl_implementation_architecture_design.pdf — implementation-phase architecture document
• docs/nehl_implementation_component_design_document.pdf — component design details
• uml/ — system context, sequence, and activity diagrams

Current expectations

Use this repository for:

• studying the project structure
• reviewing the original code and design documents
• mining SUMO/RabbitMQ multi-agent experiment ideas
• preserving the original MSE project history

Do not expect:

• ongoing maintenance
• modern dependency management
• current-version compatibility
• production readiness