/MetroMantap

Real-Time Crowd Management System for Railway Stations

Primary LanguageHTML

MetroMantap

Real-Time Crowd Management System for train stations using sensor technology, AI-driven analytics, and real-time data processing to improve passenger flow, reduce congestion, and optimize train operations and scheduling

Future Enhancements

  • Color grouping for passengers (dashboard)
  • GET request from LSTM API (dashboard)
  • Asymmetric Encryption
  • Cabin exit gesture detection
  • Use GET request to scrape ridership data (LSTM)

Server Setup - Central Processing (PC)

Pre-requisites:

Setup

  1. Install InfluxDB and extract the zip to C:\Program Files\InfluxData\, rename and move all files to parent folder

  2. Start InfluxDB

    cd 'C:\Program Files\InfluxData' && ./influxd

  3. Grant network access, InfluxDB UI can be viewed at http://localhost:8086

  4. In InfluxDB UI, setup user account with org NextGen Hackathon to obtain All Access API Token, save the token

  5. In InfluxDB UI > Load Data > Buckets, create 2 buckets main and rasp-pi

  6. Under Dashboard, Create Dashboard > Add a Template, paste this URL

  7. Open new terminal, clone the repo and navigate to root directory

    git clone https://github.com/Vincexodus/MetroMantap.git && cd .\MetroMantap\

  8. Duplicate .env.example and rename the duplicated file to .env.local, replace the values accordingly

  9. Install OpenSSL and generate self-signed ssl cert

    # root directory
cd assets && ./ssl_cert.sh

  10. Copy generated folder called ssl to Raspberry PI.

  11. Create virtual python environment and activate it

    virtualenv venv
venv\Scripts\activate

  12. Download required python modules

    pip install -r requirements.txt

  13. Run server script (recommend with PyTorch + Cuda installed)

    py pc_server.py

Client Setup - Raspberry Pi 4

  1. Install telegraf

     wget -q https://repos.influxdata.com/influxdata-archive_compat.key
 sudo apt-get update && sudo apt-get install telegraf

  2. Add environment variables for Telegraf

     export INFLUX_HOST=http://<PC-IP-ADDRESS>:8086 # PC IPv4 Address, NOT localhost
 export INFLUX_TOKEN=<ALL_ACCESS_API_TOKEN>
 export INFLUX_ORG="NextGen Hackathon"

  3. Start Telegraf, command can be found InfluxDB UI > Load Data > Telegraf > Setup Instructions

     telegraf --config http://<PC-IP-ADDRESS>:8086/api/v2/telegrafs/...

  4. Open new terminal, clone the repo and install python modules

    git clone https://github.com/Vincexodus/MetroMantap.git && cd .\MetroMantap\
sudo pip install opencv-python spidev ssl

  5. Enable SPI on the Raspberry Pi

     sudo raspi-config

  6. Navigate to Interfacing Options > SPI > Enable.

  7. Run client script raspberry_client.py in Thonny IDE

LSTM Model Training and Inference

  1. Data Collection:

    • Download the HTML file from data.gov.my dashboard
    • Run the following commands to extract and scrape data:
       cd ridership-LSTM
   python data-scraping/train-combination.py  # Extract possible OD pair combinations
   python data-scraping/od-ridership.py       # Scrape the latest daily OD ridership data
   python data-scraping/monthly-ridership.py  # Query the OpenAPI for monthly ridership data

  2. Data Preparation:

    • Use data-preparation/prepare-data.py to process and prepare the collected data

  3. Model Training:

    • model/lstm.py is used to train the LSTM models for each OD pair

  4. API:

    • api.py provides an API interface for the trained models
    • Open a new terminal and navigate to ridership-LSTM directory
    • Start the app using uvicorn api:app --reload
    • This endpoint reads the future_predictions.csv file from each OD pair folder in the model/lstm/ directory and returns the predictions as a JSON response
    • The API will be available at http://0.0.0.0:8000/predictions/

References

Acknowledgments

  • Data source: data.gov.my
  • This project was developed as part of the MMU ZTE 5G Hackathon