June 30, 2023

Advancing Satellite Autonomy: Building Trust in Data-Limited Environments

Dr Christopher Capon

Advancing Satellite Autonomy: Building Trust in Data-Limited Environments

The concept of digital twins has gained significant traction across various industries, revolutionizing the way complex systems are designed, tested, and operated. One sector that has embraced the potential of digital twins is the automotive industry, where they have played a crucial role in the development of autonomous vehicles. Interestingly, this same approach is now being applied to satellites, enabling the advancement of autonomy in space exploration.

In this blog post, we will explore how digital twins are being utilized to progress automation within the satellite industry, drawing parallels with the successful use of digital twins in the automotive domain. We also explore how Nominal Systems’ digital twin software is providing a synthetic environment for training and testing intelligent constellation operations within the University of New South Wales’ AI Testbed.

Digital Twins: A Brief Overview:

Digital twins are virtual replicas of physical assets, systems, or processes. They provide a real-time digital representation of an object's behaviour, allowing engineers, designers, and operators to understand its performance, predict outcomes, and optimize operations. By utilizing sensors, data analytics, and simulation technologies, digital twins enable precise monitoring, analysis, and control of assets, resulting in enhanced efficiency, reliability, and safety.

Autonomous Vehicles: Pioneering the Use of Digital Twins:

The automotive industry has been at the forefront of digital twin utilization. Manufacturers leverage digital twins to simulate and test different scenarios, optimizing vehicle performance, safety features, and efficiency. For instance, Tesla, a leading electric vehicle manufacturer, extensively uses digital twins to evaluate and enhance autonomous driving capabilities. By feeding real-time data from its fleet of vehicles into digital twin models, Tesla can gain large data sets of real-life driving scenarios from its on-board sensors. Tesla can also use real-time data to calibrate digital twins of their cars to produce more accurate representations of their vehicles. With this accurate real-life data they can provide an more representative digital environment to train, test and improve their self-driving algorithms, providing safer and more reliable autonomous functionality.

Applying Digital Twins to Satellite Autonomy:

Space exploration and satellite technology are undergoing a transformative phase, with an increasing focus on developing autonomous satellites capable of performing complex tasks without human intervention. Digital twins have emerged as a valuable tool in this pursuit, mirroring their successful implementation in the automotive industry.

  1. Design and Simulation: Digital twins enable satellite designers and engineers to create virtual models of satellite systems, simulating their behavior and performance in various scenarios. This allows them to identify potential issues, optimize system designs, and streamline the development process. By leveraging data from previous satellite missions, digital twins can simulate real-world conditions and predict the behavior of autonomous satellites, enhancing their functionality and reliability.
  2. Operational Monitoring and Predictive Maintenance: Similar to the automotive industry, digital twins play a vital role in monitoring the health and performance of satellites in real-time. By integrating sensor data from satellites into digital twin models, operators can continuously monitor key parameters, detect anomalies, and predict maintenance needs. This proactive approach minimizes downtime, optimizes satellite performance, and extends their operational lifespan.
  3. Autonomous Decision-Making: Digital twins facilitate the development of autonomous decision-making algorithms for satellites. By creating virtual replicas of the satellite and its surrounding environment, engineers can train AI systems to analyze data, make intelligent decisions, and perform complex tasks without human intervention. These algorithms can autonomously control satellite movements, manage mission objectives, and adapt to changing conditions, significantly reducing human intervention, and enabling more efficient space exploration.
  4. Training of AI Models: Digital twins offer a unique advantage in generating synthetic data to augment the training of AI algorithms in data-limited environments such as space. In space exploration, acquiring large amounts of real-world data for training AI models can be challenging and expensive. By augmenting limited real data with synthetic data from digital twins, AI algorithms can be trained on a more diverse and comprehensive dataset, improving their robustness, adaptability, and generalization capabilities. This approach allows for more efficient training and optimization of AI algorithms for space applications, even in situations where collecting extensive real-world data is not feasible.

Case Studies:

At Nominal Systems we have been working with UNSW Canberra Space to build an “AI Testbed” at the Australian National Concurrent Design Facility to enable the generalised development of federated/distributed learning concepts to transform edge learning applications on single satellites to collective learning across satellite constellations.

To achieve this outcome, Nominal is providing its digital twinning products to support the generation of realistic, synthetic datasets to a virtual constellation of 8 satellites models. These satellites are composed of a range of subsystems, including different sensors, actuators, EPS, communications, and are each connected to their own physical flight computer with different edge processing devices. These flight computers then command & control their associated digital satellite model to complete some objective, with virtual sensor data and system telemetry passed from the models to the devices to complete some learning objective. This data may also include information communicated via inter-satellite links.

This approach provides users with a testbed to rapidly conceive, develop and test onboard learning applications, along approaches to distributed learning of systems. The approach is general and can be applied to specific stakeholder problems and/or needs e.g. for autonomous heterogeneous constellation control.

Head node where Nominal Editor and other data can be hosted and fed to the network of flight computers.
Rack of flight computers representing individual spacecraft. The flight computers are integrated with and receive synthetic system, payload and environment data from Nominal's digital twin software. Nominal's digital twin software also simulates purely digital spacecraft to further populate the constellation.


Digital twins have proven to be transformative tools, revolutionizing various industries, including automotive and satellite technology. Drawing inspiration from the automotive industry, the satellite sector is harnessing the power of digital twins to advance autonomy in space exploration. By employing digital twins for design and simulation, operational monitoring, and autonomous decision-making, satellite operators can enhance efficiency, reliability, and safety, paving the way for more autonomous and intelligent satellites.

If you want to level-up your onboard autonomy, build trust in your systems and establish a sustainable competitive advantage, contact info@nominalsys.com for more information.


"Digital Twins: A Brief Overview" - GE Digital. [Link: https://www.ge.com/digital/industries/digital-twin]

"Digital Twins for Autonomous Vehicles" - Forbes. [Link: https://www.forbes.com/sites/alanohnsman/2020/03/09/how-tesla-architects-its-silicon-valley-smart-robot-cars/?sh=136c6a912112]

"Digital Twinning for Mars Rover Autonomy" - NASA JPL. [Link: https://www.jpl.nasa.gov/news/digital-twinning-for-mars-rover-autonomy]

"Earth Observation Satellites and Digital Twins" - European Space Agency. [Link: https://www.esa.int/Applications/Observing_the_Earth/Earth_observation_satellites_and_digital_twins]