Cybercom’s Machine Learning specialist Jouni Luoma won the AWS (Amazon Web Services) Summit Stockholm DeepRacer League competition 22.5.2019. Winning time was 8,7 seconds – that is a new EMEA record and just a few hundreds of a second behind the world record. By winning the Stockholm race, Jouni got ticket to AWS re:Invent in Las Vegas, where the grand finals will be raced in December. Good luck for the finals!
Jouni Luoma is working at Cybercom Finland Tampere office, which is known for its expertise in Cloud Services, 24/7 Hosting Services, Machine Learning as well as Design and Agile software development. Jouni is also keen to develop his competence by studying Machine Learning in University level.
“The race was really exciting, and the competition was tough. Every time there was any of the top5 developers at track it seemed that the result may change. I managed to get the winning lap early in the session but tried to get even better time all day as it seemed that I was not in safe. The excitement lasted until the final minute of the race as 3 persons of top5 were last developers on track and I had done my last run already before them. Now I am looking forward to compete in the DeepRacer League Finals in re:Invent 2019 coming December. For that race I still need to continue developing the model, as the competition will be fierce“, Jouni comments.
What is a DeepRacer?
AWS DeepRacer is a 1/18th scale race car which you can start training with reinforcement learning (RL). RL is an advanced machine learning (ML) technique which takes a very different approach to training models than other machine learning methods. Its super power is that it learns very complex behaviours without requiring any labelled training data and can make short term decisions while optimizing for a longer-term goal.
AWS DeepRacer offers an integrated simulation environment and reinforcement learning platform hosted on the AWS Cloud for experimentation and optimization of your autonomous racing models.
Training DeepRacer’s intelligence with Reinforcement Learning (RL)
Jouni used AWS SageMaker environment to create custom RL models for the DeepRacer car.
AWS DeepRacer integrates with Amazon SageMaker for reinforcement learning model training, AWS RoboMaker to provide the racing simulator, Amazon Kinesis Video Streams for video streaming of virtual simulation footage, Amazon S3 for model storage, and Amazon CloudWatch for log capture.
“Reinforcement learning means, that you teach your car with ‘try and fail’ method. You provide a reward function for the simulation / training environment. It is a function, from where the agent – agent, meaning here the DeepRacer - get rewards when it behaves in the desired way. The car is trying to achieve as many rewards as possible, and therefore learns to execute such actions that lead to largest amounts of rewards”, Jouni tells.
But how do you follow the learning process?
“On the simulator you can see, how the learning process is progressing. For example, for the first times your car is on the track, it just keeps driving out of the tarmac. You train it more, and next thing you notice the car is following the track as you wanted. After this, you train it to go faster and faster. I must say, that I’m very pleased the training with DeepRacer I made, but we still need to improve for the grand finals”, Jouni grins.
Cybercom and AWS - Strategic Collaboration and a Premier Consulting Partner
Amazon Web Services, Inc. (AWS) announced in December 2018 their fifth Region in Europe, first in the Nordics and located in Sweden. Cybercom announced already earlier in September 2018 a strategic collaboration with AWS as a format of Cybercom AWS Business Group (CABG) to help mid-sized companies, enterprises and public sector organizations accelerate their journey to the cloud and boost their digital transformation. We are one of the selected go-with partners to AWS in the Nordics.
The CABG will be focusing on leverage the benefits of the new launched AWS Europe (Stockholm) Region, delivering low-latency cloud services as well as allowing customers to store their data in the Nordics, overcoming any data sovereignty issues they might have.
DeepRacer | AWS DeepRacer is a 1/18th scale race car which you can start training with reinforcement learning (RL). AWS DeepRacer is also an environment where you can train the models in simulated environment.
DeepRacer League | The world’s first global autonomous racing league, open to anyone. Get on the track to compete online in the monthly Virtual Circuit races or in-person at Summit Circuit race events worldwide. The AWS DeepRacer Championship Cup finals at re:Invent 2019.
SageMaker | Amazon SageMaker is a fully-managed service that covers the entire machine learning workflow to label and prepare your data, choose an algorithm, train the model, tune and optimize it for deployment, make predictions, and take action.
RoboMaker | AWS RoboMaker is a service that makes it easy to develop, test, and deploy intelligent robotics applications at scale.
Amazon Kinesis Video Stream | Amazon Kinesis Video Streams makes it easy to securely stream video from connected devices to AWS for analytics, machine learning (ML), playback, and other processing. Kinesis Video Streams automatically provisions and elastically scales all the infrastructure needed to ingest streaming video data from millions of devices.
Amazon S3 | Amazon Simple Storage Service (Amazon S3) is an object storage service that offers industry-leading scalability, data availability, security, and performance.
Amazon CloudWatch | Cloud storage is a critical component of cloud computing, holding the information used by applications. Big data analytics, data warehouses, Internet of Things, databases, and backup and archive applications all rely on some form of data storage architecture.