Artificial intelligence (AI) has been on the rise in recent years due to improvements in hardware, machine learning models and cloud computing unleashing an array of AI products that promise to vastly increase the automation, accuracy and efficiency of analysis and decision making, including those related to engineering design and operations.
In the news recently has been ChatGPT, a chatbot by OpenAI that uses natural language processing (NLP) to generate responses to general user input. ChatGPT has been the fastest digital product to reach 1M users in history. But why? ChatGPT works by analyzing user input and using machine learning algorithms to generate responses that are based on its understanding of the user's needs. This can be a huge time-saver for software developers, as it allows them to within seconds get ChatGPT to return complete code, information related to the functionality of the code, and even unit tests all from textual prompts. The example below shows ChatGPT generating code to solve an astrodynamics problem based on an input prompt.
Beyond software engineering, artificial intelligence is also aiding in the operations of physical systems, such as self-driving cars, that use AI to predict the likelihood off a pedestrian crossing the road based on the pedestrian’s location, body posture, etc... Simulation is often used to develop and test self-driving cars, particularly in the early stages, when the algorithm is not mature enough to safely drive on public roads. Simulation provides a controlled and safe environment for the self-driving car to make mistakes and improves the efficiency of training algorithms by increasing the frequency that trial-and-error iterations can be made. Therefore, a generalised simulation framework optimised for scalability and efficiency would greatly benefit the ability to train AI to automate every Thing.
Nominal Systems is making it easy and effective for our customers to train AI models to automate complex systems by providing its efficient and realistic simulation capability as a safe-to-fail training environment. The value of Nominal Systems’ Nominal Digital Twin (NDT) architecture, is its emphasis on the following characteristics:
- Realism: The Nominal Digital Twin (NDT) architecture features a default library of advanced physics models to allow users to mimic a large range of sensors, actuators and environment relevant to their complex system. All default component models are vigorously tested to ensure their validity. Custom models can be uploaded and used for further flexibility;
- Variety: The NDT architecture can generate a diverse range of synthetic training data by affording the user flexibility in changing the configuration of their simulation and the ability to append error models to components to generate natural variation. This helps ensure AI systems are exposed and tested in diverse scenarios and edge cases;
- Scalability: Designed to be inherently multi-threaded and cloud-scalable, the NDT architecture allows the user to quickly and efficiently generate a large amount of training data to ensure the reliability of their automated system;
- Customisability: The modular NDT architecture enables users to write custom component models in most languages supported by the Mono framework, including: C#, Java, Python and others;
These characteristics are valued by University of New South Wales Canberra Space (UNSWCS) who are now using Nominal’s simulation products as part of a testbed for training intelligent, automated constellation scheduling. The University of New South Wales Canberra Space (UNSWCS) is a leading research group in the field of intelligent space systems. Including the design, development and operation of the M2 mission, arguably the most sophisticated CubeSats ever launched, featuring on-board intelligence and edge processing. With Nominal’s simulation software, UNSWCS are able to simulate full-satellite systems, including payload and bus, across a diverse range of satellite designs and orbital conditions. Thanks to the NDT, UNSWCS can efficiently build trust in their intelligent constellation algorithms within a safe to fail environment improving their confidence in applying them to the real world.