Skip to main content

AWS RoboMaker

AWS RoboMaker FAQs

Page topics

Simulation

Open all
AWS RoboMaker Simulation is a fully managed service that enables you to easily create simulation worlds and run simulation jobs without provisioning or managing any infrastructure.
There are multiple use cases for RoboMaker Simulation. You can use RoboMaker Simulation to test robotics applications in an automated manner and to speed up iterative development. You can use it to generate synthetic image or lidar data for algorithm development. You can also use RoboMaker Simulation to train machine learning models using reinforcement learning.

RoboMaker Simulation provides fully managed compute infrastructure for running simulations at any scale. RoboMake Simulation WorldForge makes it possible to automatically create hundreds of pre-defined, randomized simulation worlds that replicate real-world conditions without engineering investment or managing world generation infrastructure.

RoboMaker Simulation supports any simulation engine of your choice. You can connect your container-based simulation application to Robomaker and run simulations without provisioning, configuring, or managing any infrastructure. The service supports simulation world creation in WorldForge for Gazebo-compatible world formats.

A robot application in RoboMaker refers to the robot middleware (such as ROS) based application that you will run on a physical robot. You need to build your robot application source code to X86 architecture in order to work with RoboMaker Simulation.

A simulation application includes a 3D simulation world and simulator plugins that control the movement of a robot in a simulation world. Similar to a robot application, you need to build your simulation application source code to X86 architecture in order to work with RoboMaker Simulation.
You get started with RoboMaker Simulation run by creating a robot application and/or the simulation application. The applications link to your code stored in an Amazon S3 bucket. You can then create a simulation job to execute your simulation application and your robot application on RoboMaker’s fully managed infrastructure.
A simulation job is the main service resource in RoboMaker Simulation run. You run your simulation workloads on RoboMaker by creating a simulation job, which automatically provisions and manages the compute infrastructure for you. Learn more about creating a simulation job in our documentation.
A simulation job batch is a collection of simulation job requests. You can create a new simulation job batch using AWS RoboMaker's API. A simulation job batch allows you to submit multiple simulation job requests through a single API call and provides queuing capability for all submitted requests.
You can use the container tools that you are already familiar with to build and package your code into Open Container Initiative (OCI) compliant images and store in Amazon Elastic Conatiner Registry (ECR). You can use those images to create a RoboMaker robot application and simulation application.
RoboMaker supports versioning of robot applications and simulation applications so you can control which implementation your robots and simulations use. A version is a numbered snapshot of your robot application or simulation application you can create for use in different parts of your development workflow, such as development, beta deployment, and production.
Yes, you can delete a particular version of an application or all the versions of an application.
No, RoboMaker does not store your robot application or simulation application in the service. You upload your applications to your S3 bucket and refer to the S3 object path during robot application and simulation application creation.
You can configure the failure behavior of your simulation job to be fail or continue. In fail mode, your simulation job will fail, and the underlying infrastructure will be released if there is any failure during the simulation job run. In continue mode, your simulation job will fail, but the underlying infrastructure will be kept if there is any failure during the simulation job run. The continue mode allows you to further troubleshoot a simulation job failure by using the command line tool. In continue mode, RoboMaker Simulation charges will continue to incur until you terminate the job.
Restarting a simulation job can be used for quick iterative development. Restarting a simulation job is faster than creating a new simulation job because it doesn’t require resource provisioning and configuration behind the scenes. During robotics application development, you can use the restart simulation job feature to quickly test your code as you iterate through code changes.
Restarting a simulation job is used for quick iteration test and the updated simulation job runs on the same infrastructure resources provisioned and managed by RoboMaker. Cloning a simulation job uses the same configurations of the existing job to create a new job. It triggers infrastructure resource provisioning and configuration behind the scenes, which might take a few minutes.
The IAM role provides RoboMaker Simulation access to the resources you specified in your simulation job. For example, your S3 bucket for RoboMaker to write simulation logs into.
Simulation duration defines how long the simulation job should run until it gets terminated. The duration is based on wall clock time versus simulation time.
RoboMaker Simulation attempts to run your simulation job as fast as possible. You can also use the “real-time update rate” and “max step size” settings in your Gazebo world configuration to control how fast the simulation job should run. See the Pricing Page for details.
You are charged for simulation WorldForge based on the number of simulation worlds you generate and export.  You are charged for simulation run based on the duration of your simulation job and the CPU/Memory resources used by your simulation job. See the Pricing Page for details.
Your simulation job can run faster, slower than real time, or at real time. RoboMaker Simulation charges based on wall clock time so the speed of the simulation time doesn’t impact how you are charged. See the Pricing Page for details.

Simulation WorldForge

Open all
A simulation world is the virtual model of an environment such as a house. A simulation world is required for running simulations, and it includes the geometry, visual, and physics information of a modeled environment.

Building a simulation world can be difficult, costly and time consuming, and it also requires specialized skills in 3D modeling and simulation engines. While a few companies have the time and resources to create a single simulation world, building enough worlds to effectively scale simulation for workloads like regression testing and reinforcement learning is prohibitively difficult and costly.

RoboMaker WordForge makes it possible to automatically create hundreds of pre-defined, randomized simulation worlds that replicate real-world conditions without engineering investment, specialized design skills or managing world generation infrastructure.
You get started with simulation WorldForge by creating a world template, either from scratch or from one of the sample templates that RoboMaker provides. With a world template, you can then generate one or more simulation worlds by creating a world generation job. If you want to use the generated worlds in a RoboMaker Simulation job or in a local environment, you can do so by creating a world export job, which exports a specified world to your Amazon S3 bucket as a ROS package and in a .zip format.
A simulation world template defines the specifications of simulation worlds. For example, you can define the floorplan, rooms, and furnishings in a world template for generating a set of indoor residential worlds. A world template doesn’t precisely define a single simulation world. Instead, a world template can generate multiple simulation worlds with certain randomness such as random room sizes and furniture. The specifications of a world template defines the boundaries of such randomness.
A world generation job generates one or more simulation worlds from a world template. When creating a world generation job, you can specify the number of world variations across two dimensions – floor plans and interiors. For example, 2 floor plans with 2 interior variations per floor plan result in 4 (2*2) unique worlds.
A world export job exports a generated world as a ROS package and in a .zip format to your Amazon S3 bucket. You create a world export job when you need to use the generated world with a local environment or the generated worlds with RoboMaker Simulation run.