Deploy Intelligent Robotic Applications

Some of you reached out in response to my post about Amazon Web Services announcing AWS RoboMaker at re:Invent. The article you are reading now provides relatively more insight into AWS RoboMaker usage than my earlier post. Broadly, (1) enterprise companies with R&D and innovation arms focused on robotics and autonomous capabilities, (2) robotics companies, who manufacture and/or distribute robots, and (3) universities and research institutions benefit from the availability of AWS RoboMaker. Specifically enterprises that benefit from using this service include business units focused on the following specific use cases: logistics/ fulfillment; facilities/ construction; super or mega store retail; hospitality; research; energy/ oil and gas. Customers who may not be currently developing robotic applications in the cloud but are motivated to accelerate prototyping, decrease costs and time to production, and need more intelligence and security within their robot applications have the potential to benefit the most.

Robotics is undergoing fundamental change in collaboration, autonomous mobility, and increasing intelligence. By 2023, it’s estimated that mobile autonomous robots will emerge as the standard for logistic and fulfillment processes. By 2030, 70% of all mobile material handling equipment will be autonomous! The International Federation of Robotics has provided interesting insights on the state of this market and how it is evolving. Though the expected growth in the next couple of years is projected at 10X, robotic development has remained difficult and time consuming. Key pain points include requiring:

1. machine learning expertise for developing intelligent functions
2. many development prototyping iterations are required
3. many days spent setting up and configuring the development scaffolding environment
4. months for building a realistic simulation environment
5. duplication of efforts in integrating an app management system

Global open-source community support two products—Robot Operating System (ROS) and Gazebo that address some of the above issues:

• ROS is a set of software libraries and tools, from drivers to algorithms, that help developers build robot applications. It is the most widely used software framework for teaching and learning about robotics – over 16 million .deb (Linux Debian) packages have been downloaded in 2018, a 400% increase since 2014.
• Gazebo is a robust physics engine, high-quality graphics, and programmatic and graphical interfaces to help developers simulate robots.

With AWS RoboMaker all 5 issues above are collectively addressed. Developers can start application development with zero setup effort; create a RoboMaker development environment with a single click of a button; includes pre-installed RoboMaker cloud extensions and sample robotics applications; and download, compile and configure operating system, development software, and ROS automatically.

AWS RoboMaker cloud extensions for ROS include services such as Amazon Kinesis Video Streams (video stream), Amazon Rekognition (image and video analysis), Amazon Lex (speech recognition, Amazon Polly (speech generation), and Amazon CloudWatch (logging and monitoring).

These cloud extensions offload heavy compute of intelligent robotics function to the cloud and free developers from constraint of local compute resource on their robots.

All cloud extensions are written as familiar ROS packages along with detailed documentation and samples. Developers do not need any machine learning or cloud service expertise to use them. For example, a developer can download and add the Amazon Lex ROS package to their application and configure the microphone on the robot to send voice messages to it. The package will automatically create connection and make API calls to Amazon Lex and return the corresponding text messages. The developer can then, for example, use the text messages to command the robot for movement. The robot does not need to have powerful hardware to process voice data or convert it to text data locally. The developer does not need to learn about how to connect to the cloud and make API calls from the robot, or how to convert voice messages to text messages. 

Using the development environment, developers can start application development without any setup effort. Developers can go to the AWS Management console to create a development environment with a single click of a button. The environment will be up and running in minutes. The underlying infrastructure is automatically provisioned and the operating system, development software, and ROS are automatically downloaded, compiled, and configured.

Cloud extensions and sample robotics applications are pre-installed in the environment for quickly getting-started. 

Simulation is provided through robotics simulation as a service and supports large scale and parallel simulations. Developers simply upload their robotics application to an Amazon S3 bucket and then run a simulation with it. There is no infrastructure to provision, configure, or manage, and developers can run multiple simulations in parallel. The simulation scales the underlying infrastructure automatically based on the complexity of the simulation. Developers do not need to worry about infrastructure scaling and only pay for the resources the simulation consumes. The simulation also provides pre-built virtual 3D worlds such as indoor rooms, retail stores, and racing tracks. Developers can download these worlds, modify, and use them in their simulations with little to no capital or engineering resources! 

Fleet management service is integrated with AWS Greengrass and has robot registry, security, and fault-tolerance built-in. Once a robotics application has been developed and tested, a developer can use fleet management to deploy it into their robots over-the-air with just a few clicks on the AWS Management Console!

Share your thoughts on how you might use AWS RoboMaker either privately or by posting comments below.

About the Author: Madhu cherishes the opportunity to learn and collaborate. Note that what is expressed by Madhu here is of his own interest and is in no way reflective of his employer.