Whitepaper

Summary

More than 17,000 satellites are announced for launch into Low Earth Orbit (LEO) before 2027, for the purpose of real-time Earth observation and worldwide connectivity. The multitude of mission initiatives is a sign of democratization of space and induces enormous business opportunities.

In this context, LEOpowver harvests the results of the ERC Advanced Grant POWVER for the LEO market to deliver an orbit-proof software for the continuous, fully automated, profit maximizing, and power-optimal dynamic operation of Low Earth Satellite Constellations.

Thus, the LEOpowver toolchain solves the pivotal challenge in the software-driven orchestration of satellite constellations, namely the management of the severely limited and inter-dependent electric power budgets in orbit. To achieve this, the LEOpowver solution leverages:

Highly Accurate Power Models.

LEOpowver deep battery abstractions approximate the actual distribution of the state-of-charge of the on-board power storage very tightly. This enables tracking and extrapolation of the battery state with unprecedented accuracy.

Optimal Resource Utilization.

Large satellite constellations induce critical interdependencies between the individual satellites’ power budgets. LEOpowver enables the constellation-wide best utilization of communication and payload resources while provably minimizing battery depletion risks.

Machine Learning Mechanism.

In-orbit battery measurements are transmitted to ground at the earliest possible moments, where they are fed into LEOpowver to continuously rectify and improve the on-ground models of the satellites. This dynamic learning unlocks precise predictions of the batteries’ state of charge.

Receding-Horizon Scheduling.

The LEOpowver software perpetuates the computation of always safe and best-to-follow schedule for the upcoming period in a self-adaptive manner. Thus, flight plans are efficiently and safely concatenated into a persistent control over the satellite fleet.

Context

There is an increasing interest in deploying satellites for Earth observation, as well as in large-scale networks to provide worldwide data transport. This is framed in the so-called new space context, where:

• Technological advances and miniaturizations enabled significant reductions in components weight, size, price, and lead time. Recent missions based on Commercial-Off-The-Shelf (COTS) components on-board of inexpensive CubeSat platforms have reinforced interest in new service opportunities from LEO, now available to more countries, institutions, small consortia, and private companies than ever.
• Launcher’s reusability and cost reduction are shifting the traditional cost-driven space industry to a performance-driven one. Agile practices are thus being applied in space projects, as the cost of risk plummets. Rideshare missions are becoming the rule, deploying dozens of satellites at a time, placing the space market in a position of unprecedented growth.

Morgan Stanley estimates a $400 billion incremental revenue opportunity from providing internet access to under- or unserved parts of the world, yielding another $700 billion of indirect revenue opportunities through advertising, e-commerce, and social networking markets. Satellite networks are expected to play a key role in this context due to their unique and extensive coverage and capture up to $150 billion in revenues. This is incremental to an estimated $20 billion revenue opportunity in the Earth observation market and $10 billion revenue in the connected aircraft market.

Problem

In satellite and space systems, on-board communication and observation technologies induce notoriously high power demands that are generally difficult to counterbalance by solar infeed and on-board battery storage due to size and weight limitations. This makes the algorithmic problem of dynamic battery-powered satellite operation and inter-satellite communication a very difficult but important one.

For single satellite missions with multiple payload tasks on-board to operate, the default mode of operation is thus far a manual one: A human operator takes ad-hoc decisions about what task the satellite is to effectuate next. Our past in-orbit experiences with the GOMX-3 and GOMX-4 satellites have shown that already in this context there is massive room for improvement by properly modeling and analyzing the satellite’s battery, operational constraints, and orbital environment.

Indeed, the community has so far missed the benefits of fully automating the operation of a LEO satellite while respecting critical battery constraints and while optimizing the operational profit of the mission.

Furthermore, the human-operated solution does not scale to constellations of hundreds or thousands of satellites, as they are being ramped up. These business opportunities crucially hinge on efficient and automated concepts of operations. Especially in the presence of communication interdependencies, the management of constellations requires profound computational models as well as effective algorithmic techniques for administering their operation, based on a proper extrapolation of the electric power budget as part of the inter-satellite and satellite-to-ground communication design.

This is to be embedded into a domain-specific software infrastructure that seamlessly plugs in to the usage context of satellite operators: the LEOpowver solution.

Solution

The LEOpowver solution leverages crucial and matured components of software infrastructure for safe, energy-efficient, dynamic, and profit-optimal automated operation of LEO satellites and satellite constellations. In this, it harvests foundational results of the ERC Advanced Grant POWVER for an important and rapidly growing business domain. The LEOpowver value proposition is:

54928908 © skarintut / 123RF.com

LEOpowver has been validated and demonstrated in Low Earth Orbit. Our technology is indeed mature enough to fully automate the operation of networked satellite constellations. LEOpowver Technology Readiness Level (TRL) is advancing from 2/3 to 5/6.

Approach

LEOpowver is distinguished by its focus on highly accurate power modeling together with algorithms that navigate the space of possible future actions for the purpose of orchestrating the space mission optimally. To this end, LEOpowver extends and integrates the capabilities of existing toolchains.

Existing Tools. There are a number of general-purpose software tools and software libraries for mission planning, analysis, and operations in daily use in the LEO domain worldwide. All of them are orbit-proof by a multitude of successful space missions, yet they are missing highly crucial functionalities. Existing toolchains will be leveraged as part of the LEOpowver commercialization strategy, which will complement the missing features by seamlessly integrating the key LEOpowver innovation into existing business workflows.

Integration. In order to empower the existing and flight-validated toolchains, LEOpowver enjoys a set of flexible and rich software interfaces. In particular:

• Interface with Mission Model. LEOpowver interfaces with state-of-the-art mission modeling tools such as STK from AGI from which access information is obtained. Accurate orbital propagators as well as advanced observation/communication models can be leveraged.
• Interface with Customer MOC. LEOpowver delivers the battery-aware optimal decisions to the Mission Operations and Control (MOC) via existing APIs, e.g. based on json format. Furthermore, telemetry and payload data are retrieved via similar APIs.
• Interface with the User. LEOpowver offers a flexible interface to define the behavior of the model (battery thresholds, tasks weight, etc.) as well as a web app where the reported telemetry and computed schedules are displayed together with a 3D view of the mission model.

Elements

Besides the interfaces with external tools, at its core, LEOpowver takes over the task of continuously maximizing payload utilization while eliminating the risk of overstraining the power budget at any moment. While running, LEOpowver continually triggers a sophisticated decision-making procedure to deliver optimal schedules ready to be uploaded before each ground station pass. The overall process is a resourceful variant of receding-horizon scheduling, where schedules are continually re-computed based on the latest available information. Whenever the satellite passes over the ground station, the most recent schedule is offered for upload. Specifically, LEOpowver is comprised of the following elements:

• Orchestrator. The Orchestrator takes care of coordinating the availability of up-to-date information within the toolchain. It interfaces with the satellite operator’s API in order to receive the latest telemetry of the satellite. The telemetry, in particular logs of voltage and current measurements of the satellite’s battery, is fed to the Deep Battery model. And the flight plans provided by the module are transferred into the required format and delivered to the satellite operator’s API, after being checked for plausibility and feasibility.
• Deep Battery Model. LEOpowver leverages the Kinetic Battery Model KiBaM in the workflow of cost-optimal task scheduling. The KiBaM splits the battery charge into two parts, namely the available charge and the bound charge (one can think of the KiBaM as two wells holding fluid, interconnected by a small pipe). Unlike linear battery models, the KiBaM captures a number of non-linear effects of real batteries, like the recovery effect and the rate-capacity effect. Within LEOpowver, the battery model is able to learn and adapt based on the latest in-orbit measurements.
• Satellite Operation Interface. This interface bridges the gap between real satellites and the LEOpowver algorithmic engine. It serves two main purposes: First, this module collects telemetry measurements and stores them in an internal database where the Deep Battery model can access the data. And second, it sends computed flight plans to the satellite operator’s API, thereby converting the abstract schedule representation into the desired output format that can be uploaded to the satellite. The algorithmic engine harvests highly sophisticated dynamic programming and learning techniques.

A detailed description of the models and algorithms can be found in our 2020 IEEE Transactions on CAD paper.

Use Case

The LEOpowver tool-chain has been successfully implemented in the GomSpace GOMX-4 dual-satellite mission, launched in 2018.

To command the 6U satellites (GOMX-4A and GOMX-4B), LEOpowver was integrated with GomSpace’s Hands-Off Operations Platform HOOP, an automated, flexible, and scalable end-to-end satellite operation framework for commanding and monitoring subsystems, single-satellites, or constellation-class missions.

The dual-satellite GOMX-4 mission was used to validate and demonstrate the LEOpowver approach in orbit, allowing our integrated software to be distinguished as orbit-proof. Results proved the resulting machine learning approach is highly accurate, scalable, and flexible. A detailed report of the use case can be found in our SmallSat 2021 paper.

Commercialization

The LEOpowver technology offers operators of space missions the unique possibility to manage their missions in an economical, energy-optimal way. The European Research Council ERC has recognized the highly innovative approach and the market need for this technology and awarded an ERC Proof-of-Concept grant in the amount of 150,000 €, which provides the opportunity to further develop the technology and prepare it for commercial market entry.

Approach. The current development approach is user-centered at all times, which means that a software product is brought to market, in which the later user or user groups are actively integrated into the development from the beginning through various test projects and feedback loops. This ensures that software is brought to market that optimally meets the requirements of the customer. Thus, test projects and feedback loops are already planned with strategic partners from the field of LEO satellite technology to further optimize the software and develop it closely with the customer. GomSpace, D3TN, and Orbitare are among the strategic partners with great interest in the software.

We are planning for a base release of the software, that is available free of charge but restricted in functionality. Various options for full releases are currently being evaluated. Software licenses will be the core product of the commercialization activities and will be paired with customization support for end customers. In the long term, it is planned to scale the business model further. The software can indeed be adapted to other scenarios with little conceptual effort, for example, to optimize terrestrial applications such as electric mobility or drones. Battery optimization can also provide great added value in these application scenarios.

Team

Currently, the LEOpowver team consists of four experts from different areas who fully cover the required skill set for successfully maturing the software and its commercialization.

Dr. Holger Hermanns is Professor of Computer Science at Saarland University in Germany, holding the Chair of Dependable Systems and Software. Hermanns has a track record in formal verification, especially quantitative model checking with applications in energy informatics and space informatics. Multiple ERC Grants have been awarded to him and he has been inducted into Academia Europaea in 2013. He is the principal investigator of LEOpowver.

Dr. Juan A. Fraire is a researcher at the National Institute for Research in Digital Science and Technology (INRIA) in France, and associate professor at Universidad Nacional de Córdoba (UNC) and Saarland University in Germany. His research focuses on spaceborne networking and distributed applications enabled by state-of-the-art informatics techniques. Fraire is the founder and chair of the annual Space-Terrestrial Internetworking Workshop (STINT) since 2014, has co-authored more than 60 papers and the “Delay-Tolerant Satellite Network” book. Fraire participates in joint projects with world-renowned space agencies and companies.

Gregory Stock is a Ph.D. student at the chair of Prof. Holger Hermanns since June 2019, where he currently works as the main developer of LEOpowver. Before that, he performed a two-month internship at GomSpace Luxembourg, where he got hands-on experience regarding their in-house software portfolio supporting the operation of GOMX-4A and GOMX-4B. Already his bachelor’s thesis entitled “A Modest Approach to Satellite Operation” has contributed algorithmic and modeling support to the very problem domain of LEOpowver. Stock is the main person to execute the technical work within LEOpowver and will be the daily interface to the satellite engineers and operators of our industrial partners and customers.

Matthias Eßling holds a bachelor’s and a master’s degree in Economics & Law with a focus on digitalization and innovation topics from Saarland University. In addition to LEOpowver, he works as an incubation manager at IT Inkubator GmbH in Saarbrücken, where he develops IT start-ups in the pre-seed phase from the idea to spin-off maturity as an interim manager. He is responsible for the commercialization activities of LEOpowver.