SAVN - A Simulation Platform for Autonomous Vehicle Networks
SAVN is a simulation platform (in development) allowing researchers to visualize and benchmark traffic control algorithms for autonomous vehicle fleets.
The platform allows to test algorithms against real-world data such as popular origin/destination pairs and realistic traffic data (car, pedestrian, road hazards, etc.). It is also possible to use multiple algorithms together, in the same simulation.
How it works
SAVN is separated into three components:
- Website: Create, configure, visualize and benchmark algorithms
- Python framework: Use to connect an algorithm to SAVN, providing fleet information in real-time
- Server: Stores data and handles communication between the website and the algorithm
Installation
First, install the libraries required for the webserver by executing the following command within the webserver directory:
npm install
If desired, install the dummy data (including default cities):
node backend/insert-dummy-data.js
The MongoDB data will be stored in the folder webserver/backend/data/db, so make sure that it is writable.
If you want to use our sample algorithm, you also need to install the python dependencies by executing the following command within the sample_algorithms directory:
python3 setup.py install
Usage
First, start up mongod and load the database:
mongod --dbpath webserver/backend/data/db
In order to start the webserver, run the following command within the webserver directory:
npm run dev
If you want to attach our sample algorithm to a simulation, make sure that it is marked active in the frontend, and then run the following command within the sample_algorithms directory:
python3 non_colliding.py
You should then start seeing your specified objects like cars on the map in the frontend.