Code repository accompanying our work on 'Improving GNSS Positioning using Neural Network-based Corrections'. In this paper, we present a Deep Neural Network (DNN) for position estimation using Global Navigation Satellite System (GNSS) measurements. This work was presented virtually at ION GNSS+ 2021 conference. The presentation can be seen here and our slides can be viewed here

This code was developed in a conda environment running on CentOS 7.9.2009 in Sherlock, Stanford University's HPC.

To create the conda environment, use conda env create -f environment.yml

deep_gnss
|  config
|  data
|  py_scripts
|  src
   |  correction_network
   |  gnss_lib
   |  totalrecall

Our code is divided into two main parts: src and py-scripts. src contains the core functionality that our project is built on while py-scripts contains standalone python scripts for generating simulated data and training and evaluating the neural network. config contains .yml files to set hyper-parameters for the corresponding scripts and can be modified depending on your requirements. data contains example data files that our code is designed to work with.

Within src, the correction_network module defines the PyTorch DataLoaders and Network models; gnss_lib contains code that is used to simulate/find expected GNSS measurements; totalrecall defines functions and code used to simulate measurements based on a pre-determined 2D NED trajectory.

To run the train_*.py scripts, run the command python train_*.py prefix="name_of_your_experiment_here".

To run the data simulation code, run the command python data_gen.py.

Weights for trained networks can be found here

The Deep Sets model is taken from the original implementation

We also used the EphemerisManager from Jonathan Mitchell's analysis of the Android Raw GNSS Measurements Dataset (link to file)

Our coordinate analysis code is based on CommaAI's Laika repository

If you use this code in your research, please cite our paper

@inproceedings{kanhere2021improving,
  title={ Improving GNSS Positioning using Neural Network-based Corrections},
  author={Kanhere, Ashwin Vivek and Gupta, Shubh and Shetty, Akshay and Gao, Grace Xingxin},
  booktitle={32nd International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GNSS+ 2021}
  year={2021}
}

For any feature requests or bug reports, please submit an issue in this GitHub repository with details or a minimal working example to replicate the bug.

For any comments, suggestions or queries about our work, please contact Prof. Grace Gao at gracegao [at] stanford [dot] edu