id: https://api-evangelist.github.io/amazon-deepracer/vocabulary/amazon-deepracer-vocabulary.yaml name: Amazon DeepRacer Vocabulary description: Controlled vocabulary for Amazon DeepRacer API covering autonomous racing resources, reinforcement learning model states, vehicle configurations, leaderboard types, and racing personas. version: 1.0.0-alpha1 specificationVersion: 1.0.0-alpha1 operational: resources: - term: Car definition: A physical 1/18th scale DeepRacer vehicle registered to an AWS account. Cars can be assigned to fleets and loaded with reinforcement learning models for physical track racing. - term: Model definition: A reinforcement learning model trained in AWS SageMaker to drive a DeepRacer vehicle autonomously. Models encode a policy learned from reward function feedback during simulation training. - term: Leaderboard definition: A competitive ranking board where DeepRacer participants submit their trained models for evaluation against a specific racing track. Leaderboards may qualify participants for championship events. - term: LeaderboardSubmission definition: A participant's evaluation result on a leaderboard, including their model's lap times, average performance, and current rank relative to other submissions. - term: Track definition: A virtual or physical racing circuit used for DeepRacer model training, evaluation, and competition. Tracks vary in length, complexity, and dimension (REGULAR or WIDE). - term: Fleet definition: A group of DeepRacer physical vehicles managed together for an event or organization. Vehicles can be assigned to a fleet for coordinated management. actions: - term: ListCars definition: Retrieve a paginated list of all DeepRacer vehicles registered to an AWS account. - term: GetCar definition: Retrieve the configuration, fleet assignment, and metadata for a specific DeepRacer vehicle by its ARN. - term: UpdateCar definition: Modify the display name or fleet assignment of a registered DeepRacer vehicle. - term: ListModels definition: Retrieve a paginated list of all reinforcement learning models associated with a DeepRacer account. - term: GetModel definition: Retrieve the training status, reward function, and configuration of a specific reinforcement learning model. - term: DeleteModel definition: Permanently remove a reinforcement learning model from a DeepRacer account. - term: ListLeaderboards definition: Retrieve a paginated list of available DeepRacer leaderboards, optionally filtered by status. - term: GetLeaderboard definition: Retrieve the details, track requirements, and submission deadline for a specific leaderboard. - term: ListLeaderboardSubmissions definition: Retrieve the ranked list of participant submissions for a specific leaderboard. - term: ListTracks definition: Retrieve a paginated list of available racing tracks for training and evaluation. modelStatuses: - term: TRAINING definition: The model is currently being trained in a SageMaker training job and is not yet available for racing. - term: READY definition: The model has completed training successfully and can be submitted to leaderboards or loaded onto a physical vehicle. - term: FAILED definition: The model training job encountered an error and did not produce a usable model. - term: STOPPED definition: The model training was manually stopped before completion. leaderboardStatuses: - term: ACTIVE definition: The leaderboard is currently accepting new submissions and evaluating participant models. - term: EXPIRED definition: The leaderboard submission window has closed; rankings are final and no new submissions are accepted. trackDimensions: - term: REGULAR definition: A standard-width racing track suitable for typical DeepRacer vehicle configurations. - term: WIDE definition: An extra-wide racing track that allows more maneuvering room and is typically used in championship events. capability: personas: - term: ML Practitioner definition: A developer who trains reinforcement learning models by iterating on reward functions, hyperparameters, and training environments to improve autonomous racing performance. - term: Race Participant definition: A competitor who submits trained models to DeepRacer leaderboards, monitors rankings, and qualifies for virtual or physical racing championship events. - term: Event Organizer definition: An administrator who manages DeepRacer racing events, registers and configures physical vehicles, creates leaderboards, and monitors participant progress. workflows: - term: Evaluate Model Performance definition: The workflow for reviewing trained reinforcement learning models, checking training status, and selecting tracks for performance evaluation before leaderboard submission. - term: Participate in Leaderboard definition: The workflow for finding active leaderboards, reviewing track requirements and submission deadlines, and monitoring competitive rankings after submission. - term: Manage Racing Fleet definition: The workflow for registering, organizing, and maintaining physical DeepRacer vehicles including fleet assignment and vehicle naming for events. crossReference: awsServices: - term: Amazon SageMaker definition: The machine learning platform used to train DeepRacer reinforcement learning models. SageMaker training jobs generate the model artifacts referenced by the DeepRacer API. url: https://aws.amazon.com/sagemaker/ - term: Amazon S3 definition: Object storage used to store DeepRacer model artifacts, reward function code, training logs, and simulation data. url: https://aws.amazon.com/s3/ - term: AWS RoboMaker definition: The robotics simulation service used by DeepRacer to run virtual racing simulations during model training without requiring physical hardware. url: https://aws.amazon.com/robomaker/ - term: Amazon CloudWatch definition: Monitoring service used to track DeepRacer training metrics, model evaluation logs, and API usage statistics. url: https://aws.amazon.com/cloudwatch/ - term: AWS IAM definition: Identity and Access Management service used to control access to DeepRacer API operations and vehicle management through roles and policies. url: https://aws.amazon.com/iam/ standards: - term: OpenAPI 3.1.0 definition: The API specification format used to describe DeepRacer REST API operations, request/response schemas, and authentication requirements. url: https://spec.openapis.org/oas/v3.1.0 - term: AWS Signature Version 4 definition: The authentication protocol required for all Amazon DeepRacer API requests, using HMAC-SHA256 signing with AWS credentials. url: https://docs.aws.amazon.com/general/latest/gr/signature-version-4.html - term: Reinforcement Learning definition: The machine learning paradigm where an agent (the DeepRacer vehicle) learns to maximize cumulative reward through trial-and-error interaction with its environment (the racing track). url: https://en.wikipedia.org/wiki/Reinforcement_learning - term: JSON Schema definition: The schema language used to describe DeepRacer resource structures including Car, Model, Leaderboard, and Track objects. url: https://json-schema.org/