AWS DeepRacer Evo is an experience that introduces students to artificial intelligence and machine learning by controlling autonomous race cars. It provides a virtual environment in which reinforcement learning models can be trained without physical hardware and allows low-cost experimentation.
Developers can join the DeepRacer League to showcase their trained models on physical cars for real-world challenges like object avoidance and head-to-head racing, using reward functions and tuning hyperparameters to refine them further.
What is DeepRacer Evo?
Reinforcement learning powers AWS DeepRacer Evo, an autonomous racing car designed to help developers train machine learning models hands-on through its fully self-driving 1/18th scale autonomous car and 3D virtual racing simulator; while also creating the world’s first global autonomous racing league.
The Machine Learning platform (ML platform) is tailored to developers and students seeking to learn RL, train models, and put them through rigorous autonomous racing simulation. The AWS DeepRacer league provides another opportunity for practice; competitors may compete for prizes like scholarships or advancement to the Championship Cup!
Though machine learning (ML) may seem complex at first, the platform makes it easy to get started even without prior ML knowledge. All that’s necessary to get going is accessing an internet-connected computer and purchasing an AWS DeepRacer Evo device; then working remotely anywhere with Linux or Mac OSX compatible software is provided through AWS DeepRacer Evo device software.
With the Evo device, including its stereo camera and LIDAR (Light Detection and Ranging) sensor, you’ll gain more insight into reinforcement learning – an approach to machine learning (ML) which enables robots to develop computational thinking and make situational decisions rather than simply following pre-programmed rules.
This year, AWS DeepRacer is providing developers with more opportunities than ever before to win physical prizes. Alongside the traditional time trial races, head-to-head racing provides participants with more choices and excitement. Furthermore, their original DeepRacer device now costs less to help more people jumpstart ML.
The AWS DeepRacer League is open to everyone – from high school and college students, professional developers and enthusiasts. If you want to compete in it, register on its official website free of charge; once connected with an AWS DeepRacer device you can start training your model on its virtual track and competing against others from around the world for prizes and opportunities in a global autonomous racing league.
Get started with a virtual car and tracks
DeepRacer Evo was designed to enable developers to explore reinforcement learning. It includes a virtual racing environment, fully autonomous 1/18th scale race car, and global racing league for developers to train, test, and iterate reinforcement learning models on Amazon SageMaker before deploying them onto physical cars. Amazon will also add stereo cameras and LIDAR sensor technology so their physical cars can navigate more complex real world scenarios more effectively.
AWS DeepRacer Evo offers a user-friendly way to experience AI without complicated setup or expensive hardware costs. As a cloud-based platform, AWS DeepRacer Evo eliminates additional infrastructure needs, cutting costs while speeding development efforts faster and ensuring faster learning cycles for developers with various levels of machine learning expertise.
Users can use the AWS DeepRacer console to build and deploy their own models in order to create virtual race cars that will compete against others on virtual racing tracks. You can customize your virtual racing experience by altering track layout and car parameters; in addition, a graphical user interface displays performance indicators and helps track your progress.
Your race car could also utilize an object detection model to recognize objects on the track, making your race car suitable for tracking digital pets or objects with ease as well as industrial prototypes (warehousing, manufacturing). This could prove especially useful when used for fun applications (following digital pets around or tracking an object).
Reward functions you define in Python determine the performance of a race car. You can reward it for remaining close to the center of its track or driving as fast as possible on straight sections; your car may even learn not oversteer during turns thanks to these rewards!
As you race your virtual car in the AWS DeepRacer racing league, your rank and position are displayed prominently on a leaderboard for all to see. Compete across various tracks and challenges to earn rewards that you can redeem for physical prizes–including the AWS DeepRacer Evo car itself! In addition, this virtual racing environment fosters an active global developer community by encouraging knowledge sharing as well as friendly competition.
Get hands-on with RL
AWS DeepRacer Evo provides developers with an inexpensive, enjoyable way to gain hands-on experience with reinforcement learning (RL), an advanced machine learning technique used to train autonomous race cars using cloud computing. Compatible with the AWS DeepRacer 3D racing simulator, its 1/18th scale car offers developers an engaging environment to learn the RL process through building, training, tuning, evaluating and deploying their own models.
The car also supports a virtual racing event featuring object avoidance and dual-car head-to-head races, providing developers with a more challenging and realistic racing experience that helps advance their RL skills. Developers can participate in virtual community race time trials, monitor their progress on a leaderboard and earn prizes by winning races; then those models with outstanding performances can advance to the Championship Race itself.
To help developers get started with machine learning, AWS has designed an introductory workshop and course which covers an introduction to Recurrent Neural Networks (RL), with hands-on tutorials on using AWS DeepRacer racing platform for building, training and testing of models. Furthermore, an AWS DeepRacer Educator Program offers additional resources and support to educators and students within classroom settings.
Developers can get more race action by joining AWS DeepRacer League — the world’s first global autonomous racing league featuring its miniature racecar, DeepRacer Evo. Developers can join monthly virtual community race time trials hosted by this league.
Get a real-world experience
Thousands of developers worldwide have put their AI skills to the test with AWS DeepRacer, an 18th scale autonomous race car powered by reinforcement learning that uses reinforcement learning to learn to drive on physical or virtual tracks. Now with DeepRacer Evo available, developers can take their AI expertise one step further – by deploying their reinforcement learning models onto real cars for competition against each other in head-to-head races, object avoidance challenges and other events designed to challenge even experienced machine learning (ML) professionals.
This new 1/18th scale car features a stereo camera and Light Detection and Ranging (LIDAR) sensors to identify obstacles, as well as supporting a new race type that allows teams to deploy models that detect other vehicles in the vicinity and adjust its trajectory accordingly to avoid collisions. DeepRacer was developed as an open source solution, designed to advance advancements in self-driving technology by giving developers access to its capabilities.
Developers can quickly get up and running with the virtual car and tracks within minutes using our cloud-based 3D racing simulator, then quickly iterate their models without physical hardware requirements. For an immersive experience, developers can deploy their models onto physical cars for time trial races as part of global AI racing league events such as object avoidance racing or head-to-head dual car head racing as well as time trial events; those who excel will earn awards, prizes, recognition for their accomplishments while the program fosters vibrant communities while encouraging knowledge sharing, collaboration and healthy competition between members – something physical hardware cannot.
AWS DeepRacer program membership is free; physical cars cost not too much with one-time charges for AWS DeepRacer services; additional features may include sensor kits that enable vehicle tracking as well as more advanced 3D racing simulators that add an extra dimension. In the future, AWS DeepRacer may include hardware and software options designed to expand beyond racing simulation into education, robotics, supply chain optimization and manufacturing tasks–strengthening AI skills across industries and strengthening AI technologies across sectors. Virtual and in-person racing events will also create open, collaborative and competitive environments to foster development of Machine Learning technologies.
