By KJ Pittl, Solutions Architect, Amazon Web Services & Chuong Vu, Solutions Architect, Amazon Web Services
Developers, start your machine learning engines because this AWS DeepRacer event is fast approaching. The event starts with a Learning Session & Workshop kicking off on October 1st, followed by preliminary virtual qualifying racing trials taking place between October 1st-7th, and a thrilling live race day on October 8th, 2020. Find out more and register here.
But first, what is the AWS DeepRacer League?
A fun and educational virtual event, the AWS DeepRacer League is the world’s first global autonomous racing league where participants can get hands-on with Machine Learning (ML) via a cloud-based 3D racing simulator and 1/18th scale race car. The event offers an exciting and practical opportunity to learn ML and Reinforcement Learning (RL) processes through building, training, tuning, evaluating, and deploying your AWS DeepRacer model.
By registering, you will be invited to an introductory workshop that will show you how to get started with training your RL models. After that, you can build, train, and race your models in the virtual community race time trials and track your model’s performance on the race leaderboard. The top 10 models with the quickest times will then go head to head in the Final Championship Race!
Who should attend and participate in this AWS DeepRacer event?
Developers of all skill levels are invited to participate in this AWS DeepRacer event. You don’t need to be heavily technical to get involved, and students or those just starting out in your career are also encouraged to join the event. As long as you have an interest in Machine Learning and can for example, write basic formulas in a spreadsheet, then you will enjoy and get a lot of value out of this event. AWS DeepRacer is all about making Machine Learning accessible and fun. Over time, the global AWS DeepRacer community has grown and today the race generates a lot of enthusiasm, excitement, and competitiveness! There’s something about a tangible and fun car racing environment which simplifies the education process. If you’ll pardon the pun, AWS DeepRacer is the perfect vehicle for better understanding the Reinforcement Learning (RL) process in a very tangible, visual and controlled environment.
This AWS DeepRacer event is powered by AWS She Builds
Diversity and inclusion are top priorities for AWS and our partner communities. To that end, this AWS DeepRacer event is associated with and powered by AWS She Builds, an initiative which aims to build diverse tech communities in Australia & New Zealand.
The mission for AWS She Builds is to make the future of tech more diverse. The goal of this AWS DeepRacer event is to help demystify and democratise machine learning, a field often reserved for data scientists with PhDs. Our aim is to create opportunities for under-represented groups to thrive in the tech industry. Machine learning is one of the most in demand skills in our industry, and yet there remains a shortage of developers with machine learning skills. This AWS DeepRacer event powered by AWS She Builds provides a fun and tangible introduction to machine learning and reinforcement learning concepts.
In a big win for diversity, the 2019 edition of AWS DeepRacer League actually saw the first female winner, SOLA (sola@DNP from Japan), already a previous world record holder with 7.44s in the qualifiers, take home the AWS DeepRacer Championship Cup. Incidentally, that means our little AWS DeepRacer is travelling at the equivalent of roughly 100mph (or 160kph) if scaled up to a real size car!
Under the hood — what you will learn with AWS DeepRacer
By participating in this AWS DeepRacer event, you will enjoy a fun and practical opportunity to learn the Machine Learning (ML) and Reinforcement Learning (RL) process through building, training, tuning, evaluating, and deploying your AWS DeepRacer model.
We start with a simple reward function of just a few lines written in Python, rewarding your DeepRacer with points for taking the right actions such as successfully staying on the track. Your DeepRacer model uses these points to optimise its actions for achieving the best lap times in a 3D virtual environment. After that, we’ll progress to dive under the hood to see how AWS DeepRacer utilises AWS services such as Amazon SageMaker and AWS RoboMaker to train a fully autonomous model that can be deployed on a 1/18th model car.
RL is an advanced ML technique that takes a very different approach to training models as compared to other ML methods. It learns very complex behaviors without requiring any labeled training data, and can make short-term decisions while optimising for a longer term goal. Think of RL as a machine learning algorithm that learns through experience and rewards.
In the training process, Amazon SageMaker trains the RL models and AWS RoboMaker provides the 3D simulation environment. When training starts, the agent in AWS RoboMaker collects data from its exploration in the simulated environment. Then, after a specified amount of time you can set with a “hyper-parameter”, the data passes to Amazon SageMaker to train the model. When training is complete, the new model passes back to AWS RoboMaker to evaluate and collect more data. This iterative process over time creates a more accurate autonomous model. Achieving model convergence, that is, a state where more training will not optimise your model any further, depends on many variables such as actions, speed, track shape, learning rate, and so on. It’s all about learning how to read data and tune parameters, making iterative changes and reacting as you progress.
Join us for this AWS DeepRacer event and progress your Machine Learning journey!
Ultimately, this AWS DeepRacer event powered by AWS She Builds is all about having lots of fun while you learn in a diverse and welcoming environment. An autonomous racing car project is a great setting for simplifying many of the complex parameters that can exist in broader ML projects, providing students and developers alike with an opportunity for quick wins as you see your ML efforts come to life in a short space of time.
It’s really inspiring to experience the wonder of training an object to be autonomous. You can train and tune a racing car, but ultimately you cannot control exactly what it will do once it’s on the track. This is an exhilarating experience you won’t soon forget!
As mentioned, you don’t need a lot of programming experience to participate in this AWS DeepRacer event, so we encourage everyone of all genders and backgrounds to get involved and join us for a fun and educational experience.
Ready to hit the race track on 1 October, 2020? Learn more and register here today.