Share this project

Done

Share this project

Done
Making a learning experience for state-of-the-art robotics, autonomy and AI that is inclusive and accessible.
Making a learning experience for state-of-the-art robotics, autonomy and AI that is inclusive and accessible.
114 backers pledged $67,382 to help bring this project to life.

About

Duckietown: a playful road to learning robotics and AI project video thumbnail
Replay with sound
Play with
sound
  • Off
    • العربية
    • Deutsch
    • English
    • Español (España)
    • Français
    • हिन्दी
    • Italiano
    • 日本語
    • 한국어
    • فارسی
    • Português (Brasil)
    • Svenska
    • 简体中文
    • 繁體中文

$67,382

114

Duckietown is a playful learning experience for state-of-the-art robotics and AI. 

The Duckietown project started at the Massachusetts Institute of Technology (MIT) in 2016. To date, the Duckietown curriculum has been used to teach over 700 students in 10 different countries.

2017 Duckietown classes at NCTU Taiwan, ETH Zürich, Université de Montréal and TTI Chicago
2017 Duckietown classes at NCTU Taiwan, ETH Zürich, Université de Montréal and TTI Chicago

With Duckietown you learn what makes a self-driving car go while helping the duckies around in town!

Help us creating an affordable high-quality education experience with a new generation of educational robots.

 project video thumbnail
Replay with sound
Play with
sound

We are a group of dedicated researchers and makers who believe that the best way to learn robotics is by combining rigorous theory with a fun, hands-on experience.

In our Duckietown classes at MIT, ETH Zürich, University of Montreal, TTI Chicago and NCTU, students start with a box of parts and create a fleet of self-driving cars. Watch the documentary they made about us - or, the “duckumentary”.

 project video thumbnail
Replay with sound
Play with
sound

We want to bring this experience to everybody in the world. We need your help to produce our next generation of Duckiebots and Duckietown cities.

 project video thumbnail
Replay with sound
Play with
sound

If you are an instructor, Duckietown is an inexpensive “class-in-a-box” that you can use to teach your students a state-of-the-art autonomy class. Join our instructor community, and let your students interact with all the others.

Outside of a classroom? No problem - Duckietown is the best way you can learn about self-driving cars. You can enjoy building them and seeing them go - and then, if you wish, you can learn how things really work with our lectures and other educational material (we call you a "Makademic").

Machine learning and robotics researchers - Duckietown is the standard platform for the AI Driving Olympics at NIPS and ICRA, use it to participate in embodied AI robotic international competitions. We are implementing a growing number of "machine-learning" and "classical" baseline algorithms to make it easy to get started 

About Duckietown

Duckietown is an educational and research platform that was born from a class on self-driving cars at MIT in 2016. We are using it to rethink how robotics and autonomy are taught.

During the Duckietown learning experience, students build and program their Duckiebots to navigate autonomously through the Duckietown cities. Students work together to create new functionalities and improve existing ones.

All materials, including: the software that runs on every Duckiebot, the class slides, instruction sets for building and coding, homework exercises and more, are available for anyone to use.

We believe that Duckietown should have a low barrier of entry but still be useful for science and high-level education. The hardware should be as inexpensive as possible and the software infrastructure should make it easy to experiment with new ideas without worrying about the low level details.

We now want to get the Duckietown educational experience out to the world, and that starts with making the physical hardware (the Duckiebot cars and the Duckietown cities) more accessible: easier to get and easier to use. 

We plan to do so by redesigning parts, to reduce the cost, and by streamlining the distribution

Who are we?

We are the Duckietown Foundation, a group of dedicated and enthusiastic researchers and instructors who are committed to the promotion of robotics and AI education

The mission of our non-profit foundation is to provide robotics and AI learning experiences that are inclusive, accessible, and fun

Our goal is to get any person to use our materials and gain the necessary skills to understand the state-of-the-art of autonomy and contribute to getting this technology out there. 

Any funds raised by the foundation are used either to: 1) Develop compelling and playful learning experiences or 2) to disseminate robotics around the world.

Why Duckietown?

1) It's accessible - We are focused on making the barrier of entry (in terms of cost and time) as low as possible.

2) It's tangible - the best way to learn about robotics and autonomy is by playing with things and seeing what works and then understanding why.

3) It's realistic - Duckiebots use only a camera, which is a sensor used by real autonomous cars. Other small platforms typically use “toy” sensors like infrared or ultrasonic. This means that Duckiebots can tackle many of the real challenges of building self-driving cars but for a fraction of the cost of an actual one.

4) It's fun and creative - People in the hard sciences and engineering also need a way to express themselves artistically! There is a growing community around the project and some people have done amazing things that we never expected.

 project video thumbnail
Replay with sound
Play with
sound

 How does Duckietown work?

Duckiebots can do most of what full-scale autonomous cars can do. This includes:

Calibration:

Detecting the lanes and estimating their positions and orientation in the lane, as well as, of course, following the road:

Stopping at intersections, navigating through intersections, and coordinating with other Duckiebots:

 project video thumbnail
Replay with sound
Play with
sound

 Avoiding obstacles such as duckies and construction cones:

(although sometimes duckie beaks are detected as construction cones!)

 Keeping distance in traffic:

Localizing in a global map (knowing where they are): 

Planning a path to get a duckie to its destination:

and much more!

All of this is done with only a camera. For example lane following: 

First an image is captured from the camera. Then, it is analyzed for road markings. The detected road markings are projected onto the ground plane and matched against a template of what a road should look like. The location in the lane is then used by the controller to determine how to send voltages to the motors, attached to the wheels, and make the robot drive down the road.

Who uses Duckietown? 

1 - University instructors: We are building an advanced "class-in-a-box" to be taught at university level. 

With the class-in-a-box, instructors get everything they need to build their course in a fun learning environment: lecture slides, notes, homework exercises, and hardware and instructions to reproduce the demos that reinforce the class concepts on real robots.

How to get started for instructors

  • Take a look at all the resources that we have available (more coming all the time) to make building your class as smooth as possible. 
  • Sign up on our website and participate in the forums and online community. 
  • Back this kickstarter at the "Classroom kit" level to ensure that you will get your materials in time for your course.

Timeline for instructors:

  • September 2018: Graduate class "Introduction to Autonomous Vehicles" taught at select universities. Class-in-a-box still in preparatory phase. Students will participate in the first edition of the "AI Driving Olympics" (AI-DO I - scroll down for more) competition at NIPS.
  • January 2019: Official class-in-a-box versions of the following classes available: Introduction to Autonomous Vehicles (graduate level), Control Systems with Duckietown (graduate level). Students can compete in AI-DO II at ICRA.
  • September 2019: Official "class-in-a-box" versions of the following classes available: Introduction to Autonomous Vehicles (graduate level), Control Systems with Duckietown (graduate level), Introduction to Robotics (undergraduate level).

2 - "Makademics": Makademics are a fusion of “makers” and “academics”. These are people who want to learn and build on their own, outside of an educational institution, and also want a deep understanding of how things are working. Makademics can learn about robotics and AI by building their own Duckietown and using all of the course materials at their own pace.

How to get started for makademics:

  • Sign up on our website and participate in the forums and online global and regional communities. 
  • Take a look at all the resources that we have available to make your learning experience wonderful.
  • Back this kickstarter at the "Starter kit" level and wait for everything to show up at your door!
  • Consider backing this kickstarter additionally at the "Navigation expansion pack" level to play around with the whole host of existing functionalities. 

3 - Researchers: Duckietown is affordable and lets you focus on what you want to study. See here for a list of academic papers published using the Duckietown platform.

How to get started for researchers:

  • Register on our website and join our community.
  • Consider competing in the AI-DO to test your algorithms against the top researchers in the world. You can register here.
  • Back this kickstarter at the "Starter kit" level with the optional "Deep learning GPU accelerator" add-on and wait for everything to show up at your door!
  • Consider backing this kickstarter additionally at the "Navigation expansion pack" level to play around with the whole host of existing functionalities.

What is available now (in beta) and is going to be ready in January 2019?

The software enables your Duckiebot to do many things in Duckietown, and is built so you can easily modify our reference implementations to make things better, or completely replace parts of the code if you prefer.

The software is modular and individual "blocks" can be modified while maintaining other functionalities unaltered
The software is modular and individual "blocks" can be modified while maintaining other functionalities unaltered

Our "Duckiebook" documentation contains a huge database of information including demo reproduction manuals, background theory, slides, lecture videos, exercises, and much more. Keeping with our theme, this book is also open and can be modified. More is getting added all the time!

Simulators: Particularly for machine learning-based approaches, experience in a simulation environment can reduce the amount of learning that needs to happen in the real world.

Logswe have many hours of video and associated control commands from several different Duckietowns on different continents, which means that you can use this data to bootstrap their machine learning algorithms.

Hardware: Both Duckiebots and the Duckietown cities use the same hardware infrastructure and are easily programmable

Duckiebots use only a camera to perceive the world around them and operate within it. All the decision making is done on an onboard computer and communication between Duckiebots happens by blinking of lights.

Duckiebots are built from affordable components
Duckiebots are built from affordable components
Customizing parts leads to a simpler design
Customizing parts leads to a simpler design

Duckietowns are designed with rigorous appearance specifications. These design rules make it so that cities are modular, and can be assembled to create many different city layouts.

Duckiebots and Duckietowns are co-designed to create a customizable ecosystem
Duckiebots and Duckietowns are co-designed to create a customizable ecosystem
Duckietowns are modular environments for Duckiebots
Duckietowns are modular environments for Duckiebots

 What are the next steps?

We want to expand the use of Duckietown dramatically around the world. To achieve this, we need to make the Duckietown platform cheaper, easier to obtain and able to provide more learning experiences and opportunities for cutting edge research.

With the support from this kickstarter, we will focus on the platform distribution and development.

  • Platform distribution 

To improve dissemination, we need to make the platform more accessible. For us, this means making it cheaper so more people can afford it and easier to obtain to reduce the entry barrier in terms of time (and headaches) necessary to source the parts. 

To achieve both these goals, we will develop a "one-click-solution" for obtaining all Duckietown related hardware. This will substantially cut the time and effort necessary to source the components and guarantee compatibility of the hardware with the available software. 

  • Platform development

We will make the Duckietown platform cheaper, but also more capable: allowing learners to engage in more learning experiences and broadening opportunities for performing cutting edge research.

To date, Duckiebots and Duckietowns have been built with off-the-shelf components. This leads to redundant pieces and an overall suboptimality of design.

Custom design of some of the critical components will allow us to bring down the price by focusing on including only what is necessary. At the same time, designing custom components enables us to include special features at hardware level.

We are constantly developing new hardware solutions for more accessible Duckiebots and smarter Duckietowns
We are constantly developing new hardware solutions for more accessible Duckiebots and smarter Duckietowns

We envision creating a Duckietown ecosystem where fully modular Duckietowns and fleets of Duckiebots operate in a integrated environment that is completely autonomous, programmable, and expandable.

Our first step in this direction are Robotariums ("robotic aquariums"): environments where Duckiebots continuously operate in cities that are specially designed to enable 24/7 autonomous operations. Robotariums will be remotely accessible by users around the world who would like to test their algorithms but might not have access to the hardware. Prototype Robotariums are currently being developed by the organizing institutions of the AI Driving Olympics (AI-DO).

The AI Driving Olympics

The Duckietown Foundation is excited to announce the official opening of the AI Driving Olympics (AI-DO), a new competition focused on AI for self-driving cars. The first edition of the AI-DO will take place in December 2018, at NIPS, the premiere machine learning conference, in Montreal. This is the first competition ever that will take place at a machine learning conference with real robots.

The AI Driving Olympics is presented in collaboration with 6 academic institutions: ETH Zürich (Switzerland), Université de Montréal (Canada), National Chiao Tung University (Taiwan), Toyota Technological Institute at Chicago (USA), Tsinghua University (China) and Georgia Tech (USA), as well as two industry co-organizers: nuTonomy and Amazon Web Services (AWS).

The competition will comprise 5 challenges of increasing complexity, from lane following to fleet management. 

AI-DO I Competition timeline:

  • now to October 1: Open development of the competition with periodic releases of simulators, tools and baseline implementations. 
  • October 1 to December 1: Official opening of AI Driving Olympics 2018. “Robotarium” live environments available for use. 
  • December 7: Finals at NIPS 2018, in Montréal.

AI-DO II: The second edition of the AI Driving Olympics is already scheduled to take place in May 2019 in conjunction with the International Conference on Robotics and Automation (ICRA) 2019, also in Montréal.

Reward Details

We need your support. We offer several ways that you can back us with different reward levels.

Sponsorship - Join the Flock 

If you are not interested in the hardware but still would like to support our mission then we have several options for sponsorship in this category. We are happy to work with you on appropriate ways that you can be recognized as a supporter of the project based on your specific desires and level of contribution.

The Starter Kit

For people who want to get started building and running their Duckiebot and learning about robotics and AI, this is the category for you. Only a limited number of kits are available, at a reduced price, for early adopters.

The early bird kits are delivered by Oct. 2018 and can be used for competing in the AI-DO I, NIPS 2018 competition. 

The unlimited starter kit reward deliver date is scheduled for Jan. 2019 and can be used by competitors in AI-DO II - ICRA 2019.

The starter kit includes:

  • A "Duckiebox", filled with Duckiebot parts: Raspberry Pi 3 B+ and an additional board for motor and LED control, a camera with fisheye lens, a chassis, a battery, 2 DC motors, a backplate and all the other bits and pieces needed to put it all together.
  • DIY Duckietown parts: tiles for modular city construction, yellow and white duct tape, signs, other fun stuff like duckies and stickers.

What can you do with it:

The starter kit allows you to explore the curriculum up to and including "Control". This includes the topics: "Introduction", "Modern Robotic Systems", "Architectures", "Representations", "Modeling", "Calibration", "Signal Processing", "Computer Vision", "Estimation", and "Control". You can complete all of the exercises and notebooks related to this material (including computer vision, "augmented reality", "feature-based SLAM" and more). You can assemble and calibrate your robot, and you can run the "Lane following" demo and tune parameters in the various modules and see their effects.

You can also reproduce all of the submissions to the AI-DO in the lane following with static obstacles challenge with the "purist" configuration (no GPU add-on).

Add-ons:

  • Extra Duckiebot (up to reward limit for early bird - priority given to starter packs, extra Duckiebots delivered successively): One is a robotic agent, two are a small fleet. It's always better to have more Duckiebots!

What can you do with it: 

You can do onboard detection of other Duckiebots. You replicate the "Make way for ducklings" demo and make Duckiebots follow each other at a specified distance (this works even outside of Duckietown cities).

You can also reproduce all of the submissions to the AI-DO for the challenge "lane following with dynamic obstacles" in the purist category (no GPU add-on).

  • City expansion pack: More tiles and tapes, signs and duckies.

What can you do with it:

You can build bigger and funner cities with more complicated loops.

  • Deep learning GPU accelerator: One Intel Movidius Neural Compute Stick.

What can you do with it:

You can put trained neural network models onto your Duckiebot and run them. This will allow you to follow the machine learning and deep learning components of the curriculum more easily. You can do things like end-to-end imitation learning and sim2real reinforcement learning transfer. As a result, you can reproduce all of the submissions to the AI-DO for the lane following (with dynamic obstacles if you also took the extra duckiebot add-on) for the AI-DO in the "non-purist" category.

  • Assembly add-on (not available for early bird): Your Duckiebot is shipped assembled and initialized.

Navigation Expansion Pack

Now you can make complex Duckietowns with intersections, traffic lights and way more duckies! The discounted early bird version will be delivered in October 2018, and can be used for competitors in AI-DO I. The unlimited version will be delivered in January 2019 and can be used by competitors in AI-DO II.

The navigation expansion pack includes:

  • more tiles, 
  • red tape for making 3 and 4 way intersections, 
  • a traffic light, 
  • more traffic signs.

What you can do with it:

You can now follow the curriculum sections on planning. Your Duckiebot needs "stateful" representations in order to know whether it is at an intersection or not and must choose amongst a set of different controllers in order to be able to navigate in these different cases. Your robot will also have to be able to detect the traffic light, through more advanced computer vision techniques.

You also can now run all of the entries to the navigation challenge at the AI-DO in the purist category (non-purist if you got the GPU extensions with your starter kit).

Add-ons:

  • City expansion pack: More tiles, tapes, signs, and duckies. 

What you can do with it: You can build bigger, more challenging and funner cities. 

  • Extra Duckiebot: Another Duckiebot (and duckies, of course)! 

What you can do with it:

Now you have other robots to coordinate in your Duckietown. You can complete the fleet-level planning and mobility-on-demand portions of the curriculum. 

Finally, you can now run all of the entries submitted all challenges at the AI-DO in the "purist" category.

  • Extra traffic lights and signs: Bigger, more complex and more fun!

Classroom Kit

We also offer a "one-click" solution for instructors that would like to teach a class based on Duckietown. This is perfect for teaching the "Intro to Autonomy" or "Control Systems with Duckietown" graduate classes in January 2019 or later or the "Introduction to Robotics" undergraduate class in September 2019 or later.

Donations

Here at the Duckietown Foundation we are committed to the dissemination of robotics. In particular this means getting robots into the hands of people who otherwise couldn't have access to or afford them. If you would like to contribute to the project but are not necessarily interested in building or running a Duckiebot you can donate one to someone in need. We will solicit applications through the Duckietown Donation Program and we will select the most worthy candidates. They will be made aware that their robot was provided with your support. We can also recognize you in other ways that we are happy to discuss with you. We offer two levels of donation:

  • Donate a Starter Kit: A Starter Kit will be delivered to the chosen applicant.
  • Donate a Class: A Classroom kit will be donated to the chosen applicant.

We have stretch goals too! 

People love Duckietown!

Media Coverage about Duckietown

                                                                                                                                                                             

Risks and challenges

The biggest risk is that we will not be able to source and distribute all of the components in time for the early birds.

The mitigation is that they will be delivered one month later than the posted timeline.

Learn about accountability on Kickstarter

Questions about this project? Check out the FAQ

Support

  1. Select this reward

    Pledge US$ 5 or more About US$ 5

    Join the Flock

    Help us make the Duckietown experience wonderful with your support.

    Includes:
    • Your name on an official Duckietown street sign
    • (optional) fan pack (duckies and stickers) - add $20
    • (optional) commemoration as "Duckietown Supporter" - add $100
    • (optional) commemoration as "Duckietown Defender" - add $500
    • (optional) commemoration as "Duckietown Patron" - add $1000
    Less
    Estimated delivery
    25 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  2. Reward no longer available

    Pledge $149 or more About $149

    Get a Duckiebot (early bird)

    Get started with a Duckiebot! (Limited supply at strongly discounted price for super early birds)

    Includes:
    • Duckiebot parts
    • 3 Traffic signs
    Less
    Estimated delivery
    Ships to Anywhere in the world
    Reward no longer available 32 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  3. Select this reward

    Pledge $299 or more About $299

    Donate a Starter Kit

    The Duckietown Foundation will find a deserving recipient of a Starter Kit through our Duckietown Dissemination Program and deliver it to them in your name. Your donation will be recognized.

    Note that this reward has a lower cost since the Duckietown Foundation will also contribute.

    Includes:
    • A Duckietown Starter Kit donated to a student in need
    Less
    Estimated delivery
    Ships to Anywhere in the world
    3 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  4. Select this reward

    Pledge $399 or more About $399

    Get a Starter Kit (early bird)

    Get started with Duckietown!
    (Limited supply at discounted price for early bird)

    Includes:
    • Duckiebot parts
    • Duckietown parts (enough to make a small loop)
    • (optional) Extra Duckiebot (early bird) - add $239
    • (optional) Deep learning GPU accelerator (early bird) - add $129
    • (optional) City expansion pack (early bird) - add $79
    Less
    Estimated delivery
    Ships to Anywhere in the world
    Limited 21 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  5. Select this reward

    Pledge $449 or more About $449

    Get a Starter Kit

    Get started with Duckietown!

    Includes:
    • Duckiebot parts
    • Duckietown parts (enough to make a small loop)
    • (optional) Duckiebot assembly and initialization - add $99
    • (optional) Extra Duckiebot - add $279
    • (optional) Deep learning GPU accelerator - add $139
    • (optional) City expansion pack - add $89
    Less
    Estimated delivery
    Ships to Anywhere in the world
    2 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  6. Select this reward

    Pledge $449 or more About $449

    Navigation Expansion Pack (early bird)

    Expand your Duckietown with intersections, road signs and traffic lights.
    (limited number available at discounted price for early bird)

    Includes:
    • Three-way intersections
    • 4-way intersections
    • Traffic light
    • Traffic signs
    • (optional) Extra Duckiebot (early bird) - add $239
    • (optional) Extra traffic signs (early bird) - add $49
    • (optional) Extra traffic light (early bird) - add $249
    • (optional) City expansion pack (early bird) - add $79
    Less
    Estimated delivery
    Ships to Anywhere in the world
    Limited 7 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  7. Select this reward

    Pledge $499 or more About $499

    Navigation Expansion Pack

    Expand your Duckietown with intersections, road signs and traffic lights.

    Includes:
    • Three-way intersections
    • 4-way intersections
    • Traffic light
    • Traffic signs
    • (optional) Extra Duckiebot - add $279
    • (optional) City expansion pack - add $89
    • (optional) Extra traffic signs - add $69
    • (optional) Extra traffic light - add $279
    Less
    Estimated delivery
    Ships to Anywhere in the world
    2 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  8. Select this reward

    Pledge $1,999 or more About $1,999

    Get a Small Classroom kit

    Get enough Duckiebots and a simple but complex enough Duckietown to experience the full spectrum of Duckietown challenges.

    Includes:
    • Duckiebox
    • Traffic light
    • Duckietown parts (enough to make a track with intersections)
    Less
    Estimated delivery
    Ships to Anywhere in the world
    3 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  9. Select this reward

    Pledge $2,749 or more About $2,749

    Donate a Classroom kit

    The Duckietown Foundation will find a deserving recipient of a Classroom Kit through our Duckietown Dissemination Program and deliver it to them in your name. Your donation will be recognized.

    Note that this reward has a lower cost since the Duckietown Foundation will also contribute.

    Includes:
    • A Duckietown Class Kit donated to a University in need
    Less
    Estimated delivery
    Ships to Anywhere in the world
    1 backer
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.
  10. Select this reward

    Pledge $3,499 or more About $3,499

    Get a Classroom kit

    Get enough of Duckietown to start your very own class! Two Duckiebots will be by October 2018, and the rest in January 2019.

    Includes:
    • 12× Duckiebox
    • Traffic light
    • Duckietown to meet your specifications
    Less
    Estimated delivery
    Ships to Anywhere in the world
    2 backers
    $
    Kickstarter is not a store.

    It's a way to bring creative projects to life.

    Learn more about accountability.

Funding period

- (21 days)