FAQ - Duality AI

What is USD?

Universal Scene Description (USD) composition system lets users combine scene description from multiple sources (layers) into a single “composed” view. This is extremely useful for Digital Twin and scenario creation pipelines because it lets us modularly build up large assets from smaller ones. This is a significant benefit for simulation since it allows for the continual recomposing of scenarios without having to start from scratch. It also gives you the ability to perform non-destructive overrides, so you can change how an asset might look without actually changing the data in the asset itself. The reusability created by this form of augmentation significantly increases the long-term value of your Digital Twins.

What skill-sets are needed to use Falcon?

While high-level familiarity with the concepts of simulation is helpful, working with Falcon is customizable to your requirements. We can tailor Falcon to your specific use-case with a user-friendly interface. More technical users can customize their Falcon experience via Python scripting, Unreal Engine plugins, and more. Falcon's flexibility rivals that of open-source tools, while benefiting from enterprise-level support.

Is Falcon for us? Are there specific industries and applications Falcon is aimed at?

The Enterprise Metaverse and the end-to-end workflows enabled by Falcon can be leveraged for any use case. Falcon's always expanding library of validated sensors and ability model dynamics from hydro- to aero- and everything in between puts most simulation scenarios within reach. Though Falcon has shown utility in scenarios ranging from Smart Cities to wind energy, we are currently focused on the Autonomous Robotics, Government, and Consumer Packaged Goods sectors. Our customers and partners include F500 companies, cutting edge startups, and university research groups.

What makes Duality's Synthetic Data relevant and valuable?

Synthetic data is only valuable if it provides novel information that is representative of real-world scenarios and outcomes. Highly realistic and relevant synthetic data has been shown to match and augment real-world data leading to robust and deployable ML models. Since the quality of our synthetic data helps predict successful ML model training, we developed a system for quantifying the quality of our data and guide its generation via 3 interconnected criteria: Indistinguishability, Information Richness, and Intentionality. Collectively, they are referred to as The Three I's, and you can learn about them here.

What is a Digital Twin?

Digital Twins are virtual representation of real-world entities and processes, synchronized at a specified frequency and fidelity. Simply put - they are accurate versions of their real-world counterparts that possess all the same properties. For example: a Digital Twin of an autonomous robot behaves identically to its physical twin - its physics is the same and it runs the same software. This means we can use Digital Twins to develop, learn about, improve and test products in a safe, more cost-effective and efficient way. Since Digital Twins are reusable and modular (and they aren't tied to any specific simulation) we use them to create and simulate diverse varieties of scenarios, and we generate vast amounts of high-quality synthetic data that has vital uses in the real world. Since we compose simulation scenarios from various necessary Digital Twins - Digital Twins become the building blocks of any Enterprise Metaverse.

What is a Enterprise Metaverse?

Unlike the consumer metaverse, the Enterprise Metaverse is one designed for commercial, industrial, and manufacturing applications. Instead of avatars operated by people in unbounded scenarios, we use Digital Twins of items, systems, and environments that are closely tethered to their real-world counterparts with data. This makes it ideal for gathering data on machine interactions with the environment, with humans or with other machines. In the Enterprise Metaverse, we can create real-world scenarios that help us close data gaps in machine learning and behavior gaps in robotics and workflow automation.

What is a Digital Twin Integration Platform? And what is an end-to-end Digital Twin workflow?

A Digital Twin Integration Platform like Falcon centralizes every aspect of working with Digital Twin in one place. This centralization enables a true end-to-end workflow: With Falcon you can (1) easily acquire interoperable Digital Twins from most 3D model sources. You can (2) edit and manage your Digital Twins and their associated scenarios. You can (3) modularly compose new scenarios and (4) run simulations to generate synthetic perception data using customizable virtual sensors. And you can (5) store, share, stream, and annotate your data in anyway that is of most use to your project.

What is the difference between Predictive Digital Twins and Internet-Of-Things Digital Twins?

While there is overlap between the two, Predictive Digital Twins are used for simulation while IOT Digital Twins are used for monitoring and controlling concurrently happening events. In other words, IOT Digital Twins show the present while Predictive Digital Twins are used to generating synthetic data to answer question and explore complex scenarios in ways that are too difficult to effectively execute in the real world.

What is the difference between a 3D asset and a Digital Twin?

Beyond looking realistic, Digital Twins also need to be “simulation-ready”, which means they need to function like their real-world counterparts. They need to carry information and exhibit properties tied to their ontology and physics (and more). In every way possible, they must be grounded in real-world data. And this data structure needs to be accessible and interoperable. A Digital Twin can be viewed as an encapsulation of all the relevant data that the utilization of that Digital Twin may require.