UAM-Path-Planning

Paper

Real-Time Communication-Aware Ride-Sharing Route Planning for Urban Air Mobility: A Multi-Source Hybrid Attention Reinforcement Learning Approach, accpeted by ICC 2025.

Overview

This work proposed a novel Multi-Source Hybrid Attention Reinforcement Learning (MSHA-RL) framework to solve the real-time communication-aware path planning for UAM as an "air-taxi".

Demo

Case 1

*Threshold SINR = -3 dB, in Berlin, Germany*

Case 2

*Threshold SINR = -17 dB, in Detroit, USA*

Quick start:

1. Install TransSimHub

git clone https://github.com/Traffic-Alpha/TransSimHub.git
cd TransSimHub
pip install -e .  # Install in editable mode

2. Install UAM-Path-Planning

git clone https://github.com/Traffic-Alpha/UAM-Path-Planning.git
cd UAM-Path-Planning
pip install -r requirements.txt

3. Train RL

You can train the RL model with the following code:

python train_sigppo.py