Using Project-Varuna software package for data-driven modeling, planning and controls of UGVs

Here is a beginner's guide for utilizing the Project Varuna software packages for data-centric modeling of UGVs.

1. Data-processing: In order to get started with data-driven modeling with Project Varuna, we first need to prepare the datasets in a compatible format. For release version v1.0, it is recommended that the training and test datasets are set up in one of these two formats.

   1. Project Varuna Custom: The required dataset structure and pre-curated dataset for mid-scale Hunter-SE can be found in the following GitHub repository: https://github.com/project-varuna/Project-Varuna-Datasets

   2. AutoDRIVE format: This structure is an extension for the AutoDRIVE dataset format found here: https://github.com/Tinker-Twins/AutoDRIVE-Hunter-SE-Dataset. All the individual runs from the above repository are compiled and preprocessed to include throttle to command velocity mapping, and the data has been smoothed. The compiled and processed dataset can be found at: https://github.com/project-varuna/AutoDRIVE-Processed-Datasets
      

2. Using pre-trained models (optional): If you choose to use pretrained models (currently models for 1/5th scale Hunter SE are available), all you need to do is select the models instead of training a new one in the next step. The models are available in the following repository: https://github.com/project-varuna/Hunter-SE-MMPK-Models.git
   

3. Modeling and deployment with Project Varuna software kit:

   With the datasets set up according to the aforementioned formats, use the GUI and the compiled package for modeling the dynamics, setting up parameters for motion planner and linear Model Predictive Controller. Finally, select the reference trajectory and off we go! Detailed guidelines and package release can be found here: https://github.com/project-varuna/Project-Varuna-Autonomy-Package