https://github.com/cdvel/hlea4tc
Multiagent communication artifact, with JADE and JNI (Hierarchical Reinforcement Learner)
https://github.com/cdvel/hlea4tc
research
Last synced: 3 months ago
JSON representation
Multiagent communication artifact, with JADE and JNI (Hierarchical Reinforcement Learner)
- Host: GitHub
- URL: https://github.com/cdvel/hlea4tc
- Owner: cdvel
- Created: 2015-05-03T03:32:02.000Z (about 11 years ago)
- Default Branch: main
- Last Pushed: 2025-05-28T05:42:09.000Z (about 1 year ago)
- Last Synced: 2025-08-19T03:35:38.364Z (10 months ago)
- Topics: research
- Language: Java
- Homepage:
- Size: 90.8 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# HLEA4TC - Hierarchical Learning Engine for Adaptive Traffic Control
A JADE-based multi-agent framework designed for distributed traffic control optimization with support for hierarchical agent organization and external traffic simulation integration.
## Overview
HLEA4TC provides a foundation for implementing hierarchical reinforcement learning algorithms in traffic control systems. The framework establishes a two-tier agent hierarchy where Sector agents manage groups of Junction agents, enabling coordinated traffic optimization across urban networks.
## Architecture
### Agent Types
- **Junction Agents**: Represent individual traffic intersections
- Subscribe to sector agents for coordination
- Receive state updates from external traffic simulation
- Maintain local traffic state information (coordination, priority, traffic counts)
- **Sector Agents**: Manage groups of junction agents
- Handle junction subscription requests
- Negotiate with other sectors for junction assignments
- Coordinate junction behaviors within their sector
### Key Components
- **PlatformMediator**: Central management interface for the JADE platform
- Manages agent lifecycle (creation, initialization, updates)
- Bridges between external systems and the agent platform
- **ControllerNativeInterface**: JNI bridge for simulation integration
- Connects to external traffic simulation (e.g., PARAMICS)
- Enables real-time state updates from simulation environment
### Communication Protocols
The system implements FIPA-compliant protocols:
- **Subscribe/Inform**: Junctions subscribe to sectors for coordination
- **Propose**: Sectors negotiate junction assignments with each other
## Features
- **Dynamic Agent Discovery**: Agents discover each other via JADE's Directory Facilitator
- **Hierarchical Organization**: Two-tier structure for scalable traffic management
- **External Integration**: JNI interface for connecting to traffic simulators
- **Asynchronous Updates**: Object-to-Agent (O2A) communication for real-time state updates
- **Distributed Negotiation**: Sectors autonomously negotiate junction assignments
## Current Implementation Status
The framework currently provides:
- ✅ Multi-agent communication infrastructure
- ✅ Hierarchical agent organization
- ✅ External simulation integration via JNI
- ✅ Dynamic agent management
- ✅ Basic negotiation protocols
Supports simulation-based optimization with:
- Reinforcement learning algorithms
- Q-learning or policy gradient methods
- Reward/cost functions for optimization
- State-action space definitions
- Learning rate and exploration strategies
## Usage
### Starting the Platform
```java
// Initialize platform with junction IDs
PlatformMediator.startJadePlatform("101 102 103 104");
// Add sector agents
PlatformMediator.initSectorAgent("S-001");
PlatformMediator.initSectorAgent("S-002");
```
### Updating Junction States
```java
// Update junction with traffic data
JunctionUpdateBean update = new JunctionUpdateBean(junctionID);
update.incomingCounts = Arrays.asList(10, 15, 8, 12); // N,S,E,W
update.priority = 45;
update.coordination = 0; // E-W coordination
PlatformMediator.updJunctionAgent("101");
```
## Dependencies
- JADE 4.3.1 or higher
- Java 1.6 or higher
- Native traffic simulation library (via JNI)
## Future Development
This framework is designed to support hierarchical reinforcement learning algorithms such as:
- Hierarchical Q-learning for sector-level coordination
- Multi-agent actor-critic methods for junction control
- Transfer learning between similar traffic patterns
- Coordinated exploration strategies for system-wide optimization
## License
[License information to be added]
## Contributing
[Contribution guidelines to be added]