Self-Driving Cars with Duckietown: Learning Autonomy on the Jetson Nano | Hacker News Reader

Self-Driving Cars with Duckietown: Learning Autonomy on the Jetson Nano

duckietown.org

124 pointsArtWomb38 comments

Having taken this class in person, I can attest that Duckietown is an outstanding introduction to autonomous vehicles. It covers many advanced topics from perception, filtering, and localization, to planning and vehicle locomotion - thousands of hours went into the technical content, and the platform has some great documentation, infrastructure and tools. The best part is the hands-on approach, participants contribute to the Duckietown platform and compete in a semi-annual miniature DARPA-style challenge. If you’re looking for a practical introduction to self-driving, there really is none better IMHO.

Are there parts list/guide to construct the vehicle ourself using Jetson Nano and off-the-shelf parts?

Not for this version. The Jetson powered Duckiebot has several components that are custom designed: chassis, electronics, and even the battery!

Shipping hardware instead of a simulation is the killer move here. It really distinguishes this class from others you can take at home.

I would be curious if any other course pairs their offering with hardware. The big one is Udacity and I cannot tell from their syllabus if they have hardware. (They don't mention any, so I assume they do not).

To my knowledge, this is the first MOOC on self-driving cars where each student can have their own robot home to play with.

Our feeling is that to really learn about robotics you need to touch and see things with your own eyes. This was one of the major motivating factors for us.

By the way, because it's a kit that you assemble, after you finish this course (and the follow-ups to come out) can do other projects with the Jetson Nano, sensors, and servos included.

I am Andrea Censi, one of the creators (https://censi.science). Feel free to AMA!

I'm interested in the student prereqs, and wondering if you could elaborate on them.

You've got four items on there:

  - Basic Linux, Python, Git

  - Elements of linear algebra, probability, calculus

  - Elements of kinematics, dynamics

  - Computer with native Ubuntu installation

It's clear enough what the first and last item demand of the student, but the rest, I'm not so sure. edx lists the course as "Introductory" and I guess the question is, "for whom?" :-)

Hi, If you have taken introductory courses on linear algebra, probability, calculus, and mechanics, all of these will prove helpful, but should be considered as required. We will try our best to provide all the basic knowledge required to understand the concepts.

Good point, probably we should describe them in more detail.

For the second point ("elements of linear algebra .. ") we are going to need to use matrices to represent coordinate systems.

For the third point ("elements of kinematics, dynamics") we are going to write down the equation of motions of the robot with time derivatives (ODEs).

If you have never heard of these things you might want to read up on them first.

That's great information. Thank you for your reply. (and to liampaull as well)

I have experience in deep learning research/engineering in NLP and am getting into ml for robotics.

Really excited for this course + hardware. I think one top value for me personally is that more time could be spent on hacking the ml algorithms, replicating and comparing with real research work (https://www.duckietown.org/archives/63665).

I live in India and find the shipping costs of MOOC Founder's Edition Kit (369 USD + 49 USD) a bit expensive. Just the Duckiebot MOOC Founder's Edition costs 14 USD to ship.idk what custom duty would the govt put on it :(.

1. Is there any alternative to shipping it? 2. Is there anyway to build similar bot from parts for the course? 3. I guess, since you're non-profit it's made such that it's cost effective rather than for profit. It would be great if you describe why you chose each of the component for the current bot.

Hi,

Unfortunately the shipping costs are a pain, we hear you, and custom duties are a thing to consider.

We are actively working to provide a much more effective shipping solution to gradually more countries, and India is high on our list. You can sign up to our mailing list or follow us on Facebook, Linkedin, Twitter or Instagram to stay tuned with updates! (https://www.duckietown.org/contact)

The more we iterate the Duckiebot design the more we tend towards customization; to better engineer it, reduce costs, and provide more engaging learning experiences to a broader audience.

The Founder's edition Duckiebot is designed to host several sensors (IMU, ToF, Camera, wheel encoders), has 4 LEDs, a screen, a special duckie-battery and a PCB that makes it easy to manage everything.

I have a DonkeyCar [1] with a Jetson Nano 4GB. Is it possible to follow the course with that hardware platform?

[1] https://www.donkeycar.com/

Hi BioGeek,

not sure, we never tested the Duckietown stack on a Donkeycar.

Seeing it from a graduate course in MIT (2016) to now, Duckietown is doing truly amazing! If you are interested in launching a Duckietown in your own community, please contact us info@duckietown.com

I'm Liam Paull, one of the other co-creators. Thank you for all the interest. We are very excited about the partnership with NVIDIA and getting this hands-on MOOC experience out into the world.

Is the navigation starter pack available separately? Getting a 404 error. Any plans to have ROS bindings for your hardware? Finally, any plans to integrate other types of hardware into the course? (e.g. I have a sphero rvr, realsense and jetson agx).

Hi Brutus,

Duckietown is based on ROS. You can assemble the robot and control it easily from ROS on your laptop (note that the stack runs on board completely).

For new hardware: we have in beta a Drone version of Duckietown for the high school: https://sites.brown.edu/duckiesky/ The Drone MOOC will arrive in time.

Hi Brutus,

thanks for pointing out the 404, it should be fixed now.

The city track for the MOOC is currently bundled with the Duckiebot in the MOOC starter kit (https://get.duckietown.com/collections/starter-kits/products...). You'll need both to follow along in the course.

Looks like this class is already shaping up to be a huge crowd favorite! Thanks to the Duckietown team for organizing the MOOC. Certainly this constitutes a major influence for the next generation of AI & Robotics enthusiasts ;)

Since you have a few years experience hacking on the Jetson Nano platform. I am wondering if you have other recommendations regarding peripherals for it? What are some must have add-ons? What is missing from the ecosystem that you would like to see?

Hi ArtWomb,

thank you! We are very excited to bring our vision of learning autonomy to the world, and it's great to see this interest.

The MOOC Duckiebot is the first of the family to be powered by NVIDIA. Our other Duckiebots as well as the Duckiedrone are Raspberry Pi based.

This said: as for peripherals you definitely want to get a compatible wifi dongle and provide cooling to the Jetson, e.g. with a fan. Add-ons (to the robot): we went for a modular, plug-and-play sensor suite and more features to enhance the hands-on experience (i.e., LEDs, screen).

Is there a simulation environment?

I've been trying to learn more about PyTorch and Tensorflow for autonomous driving. Is this the right place to start ? Are we expected to know a lot beforehand ?

Hi, In my biased opinion, this is a great place to start! You aren't required to know a lot beforehand. Basic python programming and some familiarity with Linux will be really helpful.

Duckietown led me to the world of robotics and makes me a researcher who I am now. A truly wonderful platform to learn and perform research.

I saw that being featured on Edx and it caught my eye. I'll probably take it at some point next year.

This project is really great cause AI and robotics are the most beautiful disciplines :)

Ah the ducks are so cute! What is the knowledge requirement to get started?

Duckietown seems to be great. Where could I find the starter kit?

Hi,

You can acquire the hardware through https://get.duckietown.com/

Thanks!

man I'm really looking forward to attend this! love the community around this project and I'm actually hoping to create a similar one in my hometown too (Cochin)

Please check our "class in a box" option at duckietown.org. For special request, please email us info@duckietown.com

That seems to be a really awesome project! Cannot wait for it!

Is there a simulation environment?

Hi, There is a simulation environment. Much of the work we do is done in simulation first before it is tested on the robot. For an example you can take a look at the AI-DO challenges page: https://challenges.duckietown.org/v4/ where you can see some submissions being evaluated in the simulator.

Yes, we have a simulation environment.

You can see it on https://challenges.duckietown.org/v4 , the challenges server for the AI Driving Olympics: https://aido.duckietown.org

So interesting!

very very interesting

Reply on news.ycombinator.com