Satellite mission operations using artificial intelligence on AWS

Mission operation centers (MOC) have been a critical element of humanity venturing into space, ranging from journeying to the moon in the 1960s all the way to present day and launching new satellite constellations. Traditional MOCs are composed of many different subsystems, including command and control, mission planning, flight dynamics, and payload processing.

Today’s satellite constellation operators find that, as their constellations grow, the complexity of their MOC increases. Fixed compute resources may not scale to accommodate hundreds of satellite entities and costs can spiral when factoring in additional power, space and cooling. However, customers who use cloud-based managed services leveraging Amazon Web Services (AWS), see opportunities for modernization of MOC subsystems. Additionally, moving to a cloud-based architecture provides opportunities for advancements like artificial intelligence (AI), automation, and digital twins.

The ability to plug in partner solutions for one or more cloud mission operations center (CMOC) sub-systems is critical. Cognitive Space is a leading AWS Partner delivering intelligent automation to satellite constellation operations using the CNTIENT platform. The system uses AWS-powered AI decision making to handle highly complex and dynamic satellite tasking requirements, and demanding mission requirements.

This blog post provides technical guidance for building and operating MOCs on AWS. We deep dive into two of the subsystems – scalable flight dynamics using serverless compute, and mission planning operations with CNTIENT.Optimize AI models running on Amazon Elastic Kubernetes Service (Amazon EKS). Finally, a set of actions are recommended to enable your business to build a CMOC strategy.

Background

For satellite operators, the MOC is the core functional component that enables business objectives. It contains all elements required for driving nominal and contingency operations.

Key foundational elements of a MOC include: –

• Command and control – Commanding the satellite and ingesting telemetry
• Flight dynamics – Orbital maneuvering and collision avoidance
• Mission planning – Contact planning and tasking
• Payload processing – Creating data products from raw data

Figure 1. Mission operation subsystems.

AWS and Partner CMOC solutions enable satellite operations at scale. Architected as a set of extensible microservices, one or more foundational elements can be deployed securely as infrastructure as code (IaC), with the option to mirror systems in a separate AWS Region for resiliency or compliance purposes. Operators can then take advantage of AWS analytics and AI and machine learning (ML) tools such as Amazon QuickSight and Amazon SageMaker to detect anomalies, offer predictive analytics, and provide situational awareness.

Cognitive Space offers turn-key solutions for mission planning with CNTIENT.Optimize. Using AWS ML services, the platform balances customer order priority, fleet, spacecraft, and system constraints to optimize collection planning and link management. In doing so, it frees mission operators from collection planning tasks so they can oversee the constellation at a fleet level.

Satellite missions on AWS

An example cloud-based architecture for a MOC on AWS is shown in Figure 2. The various subsystems are separated and modular, allowing flexible replacement and drop-in capability of partner or open source solutions.

Figure 2. Cloud mission operations center (CMOC) architecture on AWS.

This architecture introduces two new foundational elements:

• Data engineering subsystem – Customers can gain tremendous value by using AWS AI/ML for anomaly detection, forecasting components’ mean time between failure, and telemetry trend analysis. It starts with a structured data-lake in Amazon Simple Storage Service (Amazon S3), driven by AWS Lake Formation and AWS Glue to crawl the mission schema. Alarms are invoked through Amazon Simple Notification Service (Amazon SNS) for off-nominal telemetry.
• Orchestration, automation, and observability – Event-driven workflows with Amazon EventBridge allow rapid rescheduling of tasks based on operational changes. Automation is a major CMOC value-add with many operators standing up secondary systems using AWS CloudFormation in a different AWS Region for disaster recovery or compliance reasons. Finally, observability is key – Amazon Managed Grafana or QuickSight enable you to query, visualize, and alert on mission telemetry.

Let’s dive deep into two of the main MOC subsystems – Flight dynamics and mission planning.

Flight dynamics with serverless compute

The flight dynamics subsystem (FDS) typically requires extensive computational resources for orbit determination, maneuvering, conjunction analysis and collision avoidance. AWS Lambda can be a cost-effective choice for stateless calculations, although Amazon EKS is often used if customers have an existing Kubernetes footprint. This scalability enables customers to simulate and operate complex missions efficiently; whether they involve commanding large constellations of satellites or finding optimal maneuvers to save fuel and ensure spacecraft safety.

Figure 3. Simulating intersatellite links between low Earth orbit (LEO) and medium Earth orbit (MEO) constellations and ground terminals.

The FDS can tune the model fidelity used in its calculations to either increase speed or accuracy depending on the need. AWS serverless technologies allow customers to cost-optimize by elastically scaling resources out to meet demand and then terminating them after calculations have completed, provisioning only the required amount of compute.

CNTIENT intelligent mission operations

CNTIENT.Optimize mission planning is designed to help commercial and government satellite operators manage their fleets more efficiently. It uses AI to automate Earth observation tasking and improve business key performance indicators (KPIs) while satisfying technical constraints. Built on AWS, CNTIENT.Optimize can help optimize fleet operations at scale.

Operators are challenged to onboard new satellites quickly and handle complex and dynamic order stacks swiftly. In many cases, as shown in Figure 4, the AI engine is more efficient than traditional scheduling methods. Here, the AI prioritized high-value targets in a complex scenario and outperformed a traditional, rule-based approach by a factor of four. More acquisitions of high priority targets ultimately mean more revenue for a remote sensing constellation.

The model considers various factors when creating a schedule, such as balancing priorities, resource limitations, and operational constraints, while maximizing business outcomes. This optimization applies not only to individual satellites but also to entire constellations.

Figure 4. In this comparison, the AI engine on the right side, identifies four times more Earth observation targets than a traditional rule-based approach.

In addition to remote-sensing tasks, you can automate and optimize critical tasks such as inter-satellite communication links and ground downlink operations across hybrid constellation architectures. These optimizations are tailored to an operator’s specific needs through user-defined parameters.

Cognitive Space has successfully simulated operations for constellations with hundreds of satellites managing thousands of remote-sensing targets and communications needs. As illustrated in Figure 5, the platform uses several key AWS services. Amazon EKS and Amazon Elastic Compute Cloud (Amazon EC2) Auto Scaling form the foundation, automatically balancing compute resources based on the constellation size and mission complexity. Additionally, CNTIENT.Optimize uses managed services from Amazon Relational Database Service (Amazon RDS) for mission parameters, Amazon ElastiCache for fast response time, Amazon S3 for mission storage needs, and Amazon CloudWatch for monitoring. This combination ensures reliable, real-time decision-making for mission implementation within the secure and scalable AWS Cloud environment.

Figure 5. The CNTIENT.Optimize AWS architecture highlighted in this post.

Recommendations

It can be tricky to know where to start when evaluating a cloud-based MOC system. You can build it out yourself, or you can leverage aerospace and satellite Professional Services and AWS Partners. Professional Services help customers build cloud native, flexible CMOC systems deployed using proven AWS patterns with IaC. A digital twin environment is also included allowing for operator training and scenario planning prior to making satellite or constellation upgrades.

Engaging with AWS Partners such as Cognitive Space provides new opportunities for intelligent mission planning. Solutions such as CNTIENT.Optimize seamlessly scale up cluster nodes as the mission compute needs expand.

Summary

This walkthrough explores how to adopt a cloud-based mission operations system. Satellite fleet tasking can be optimized with AI solutions, and customers can improve the resilience and scalability of their MOC with AWS ProServe and Partner solutions.

For more aerospace and satellite learning resources, visit the AWS for Aerospace and Satellite homepage.