HandoverOptimDRL

A framework for developing and evaluating adaptive handover algorithms using deep reinforcement learning (PPO).

---

Description

HandoverOptimDRL is a framework designed to facilitate developing and evaluating handover algorithms using deep reinforcement learning, i.e., proximal policy optimization (PPO). It provides tools and environments to simulate the 3GPP handover protocol and to train and evaluate a PPO-based handover protocol.

This repository contains the source code, datasets, and trained PPO model for the paper: A Deep Reinforcement Learning-based Approach for Adaptive Handover Protocols, see reference [1].

---

Installation

To install the HandoverOptimDRL package, follow these steps:

1. Clone the repository:

   git clone https://github.com/kit-cel/HandoverOptimDRL

2. Navigate to the project directory:

   cd HandoverOptimDRL

3. Install the package:

   python -m pip install .

   i.e., to install it in editable mode/develop mode:

   python -m pip install -e .

You are now ready to use the HandoverOptimDRL framework for your projects.

---

Getting Started

Generate New Datasets

1. Generate GPX Traces: By using the open source software Simulation of Urban MObility, you can generate new user traces:

   sumo-gui -n map.net.xml -r routes.rou.xml --additional-files vehicles.rou.xml, map_buildings.poly.xml

   The GPX traces can be extracted using the included Python script:

   python -m gpx_extraction

2. Generate RSRP and SINR Traces: New RSRP and SINR traces can be generated using the open source Vienna 5G System Level Simulator developed and provided by the TU Wien. The corresponding launcher and scenario files for our simulations can be found in the matlab folder of this repository. To run the simulation, first adapt the launcher files based on the SUMO GPX files, and then start the runSimulation.m script.

Train a New PPO Agent

1. Train a PPO agent:

   python -m run train_ppo

Run Protocol Validation

You can validate the PPO-based and 3GPP handover protocols using the run.py file:

1. Validate the PPO-based protocol:

   python -m run validate_ppo

2. Validate 3GPP protocol:

   python -m run validate_3gpp

Citation [1]

If you use HandoverOptimDRL in your work, please cite our paper (full-text on IEEE Xplore):

@INPROCEEDINGS{10949111,
  author={Voigt, Johannes and Gu, Peter J. and Rost, Peter M.},
  booktitle={2025 14th International ITG Conference on Systems, Communications and Coding (SCC)}, 
  title={A Deep Reinforcement Learning-Based Approach for Adaptive Handover Protocols}, 
  year={2025},
  pages={1-6},
  keywords={Cellular networks;Base stations;Protocols;5G mobile communication;Source coding;Handover;Radio links;Timing;Reliability;Research and development;Handover;Communication Protocols;Mobility Management;Deep Reinforcement Learning},
  doi={10.1109/IEEECONF62907.2025.10949111}
}

---

License

This project is licensed under the MIT License. See the LICENSE file for details.