DonkieTown. A low-cost experimental platform for research on Automated and Conected Vehicles

[L4rralde]

DonkieTown

Imagine roads with non-Intelligent and Intelligent vehicles sharing data such as their position, velocity, sensor data, maneuver intentions and so on. Sharing this information will enhance traffic management and safety in multiple scenarios. DonkieTown is a testbed consisting of roads, an absolute localization system and Differential Drive Robots named Asinus Cars.

Asinus Cars comprises Jetson Nano developer kits. Source code is available in GitHub. They can navigate, communicate maneuver intentions (and more) and detect white donkey plush toys using a Mobilenetv1+SSD model retrained with a hand-labeled dataset of DonkieTown’s plush toys.

Hitherto, DonkieTown has been tested in multi-car scenarios such as convoy formations. Also, we have used DonkieTown in ROS Workshops targeting undergraduate students in the Mexican state of Zacatecas.

You can watch more videos on my youtube channel

[L4rralde]

This is a working example of the Neural Network that I trained for DonkieTown.

There are more features and news about my project, but I’m waiting for the best moment.

Thanks!

[L4rralde]

Finally, DonkieTown’s paper was published in IEEE Latin America Transactions in last June and I’m the first author. I put much effort on this :)

https://latamt.ieeer9.org/index.php/transactions/article/view/7756