How to Train Your Self-Driving (Dragon) Car

November 18th @ 16:00 - 19:00 (GMT +4) // HITB CommSec Track


Skill Level








Artificial Intelligence, machine learning, and autonomous vehicles are some of the hottest areas of research and business today. Talk about self-driving cars, and the names Uber, Tesla, Waymo, Volvo and NVIDIA come to mind. 

Employing a massive number of Machine Learning and Deep Learning techniques to enhance their products, models and autopilot capabilities, these companies are at the forefront of autonomous vehicle technology, deployment, and adoption. 

Donkey Car is an open source hobbyist project powered by volunteers with a shared interest to build their own self driving car and is currently one of the most popular self-driving car repositories on Github.  

Utilizing high-level self driving libraries written in Python, Donkey was developed with a focus on enabling fast experimentation and easy contribution. Built on Raspberry Pi and powered by a simple convolutional neural network (CNN), Donkey Car is the standard hardware car that most people build first. The parts cost about USD250 to USD300 and take 2 hours to assemble however you can do everything you’d normally do in The Donkey Gym! An OpenAI virtual environment for you to build, test and deploy your AI networks. 


We recommend you read the docs before the lab so everyone is more or less up to speed

Donkey is the standard car that most people build first. The parts cost about $250 to $300 and take 2 hours to assemble but in this lab you will only require to install Docker on your laptop / computer.

The Donkey Gym project is a OpenAI gym wrapper around the Self Driving Sandbox donkey simulator (sdsandbox).

The simulator is built on the the Unity game platform, uses their internal physics and graphics, and connects to a donkey Python process to use your trained model to control the simulated car.

How to ‘teach’ your car to drive

How to take your  ‘sample driving’ and build your own neural network

How to deploy your trained model

At the end of the lab you’ll get a chance to test your trained model against other participants and we’ll see who’s model scores the best lap times!


Share on facebook
Share on twitter
Share on linkedin
Share on email
Share on whatsapp

who should attend?

Anyone interested in artificial intelligence, machine learning and self-driving technology. Hardware tinkerers and hobbyists who have experience in remote control (R/C) scale cars. Students looking to get into AI or ML.


Gaining a basic understanding of self-driving car technologies and how to capture, train and deploy machine learning models and how to build your own self-driving ‘technology stack’ and hobbyist project Donkey Car.


Have the fastest time in the virtual race at the end of the lab and we’ll send you a brand new Donkey Car kit worth USD300!


Community Leader, DIYRobocar Hong Kong / Co-Founder at Hong Kong Society of Autonomous Model Vehicles

Jonathan Tse

Jonathan Tse is the project maintainer of the open source Donkey Car project. He is also the founder of, the official store selling Donkey Car Starter Kit for people who want to build their own self-driving car. He believes that AI education should be for everyone and AI should be made easier to learn for students. He is actively promoting Donkey Car to K12 schools and designed a curriculum suitable for K12 students.

Founder / CEO, Hack In The Box

Dhillon '@l33tdawg' Kannabhiran

Dhillon Andrew Kannabhiran (@l33tdawg on Twitter) is the Founder and Chief Executive Officer of Hack in The Box (, organiser of the HITBSecConf series of network security conferences which has been held annually for over a decade in various countries including Malaysia, The Netherlands and the UAE.

Prior to quitting his day job to lead the HITB team on crazy adventures around the world, Dhillon started off at the height of the dotcom craze as a technology journalist with PC World, ZDnet, MIS Asia and CNet. When the bubble burst, he moved on to a Malaysian telco as Chief IT Officer to spend his days in the world of Cisco AS5300s, in a land of packet switched networks at a time when Asterisk did not just mean ‘*’