Cédric Pralet is a researcher at ONERA, the French Aerospace Lab, where he conducts his research in the Information Processing and Systems Department. Specializing in decision problems, Cédric focuses on developing sophisticated algorithms and models to optimize resources and meet complex objectives. His work involves tackling challenges that demand innovative solutions, particularly in the field of satellite optimization and Earth observation.
What’s a decision problem?
“In a decision problem, you have a set of decision variables, that is, the degrees of freedom of the problem. Over these variables, you have a set of constraints, such as, if you consider a system like a satellite, you have memory constraints, energy constraints, and so on. You also have objectives you want to optimize, like acquiring images as quickly as possible or prioritizing specific requests. Our goal is to build a solution, a set of decisions, that optimizes the satellite’s operations while satisfying all the constraints.”
What is a typical decision problem you would find in Earth observation?
“Typically, in Earth observation, the mission center receives a set of requests from users wanting to observe different parts of the Earth. The challenge is that it’s not always possible to fulfill all requests due to resource limitations. The mission center must select which requests to prioritize while ensuring that the satellites have sufficient capacity to handle the selected tasks.”
Could you give us a practical example of a typical use case in large area coverage for Earth observation?
“For instance, some users might request images of an entire country like France, while others might specifically want a city in Italy or a region in Spain. We have to decide which areas to prioritize for the next satellite pass. For example, covering a large area like France might require several weeks of satellite passes, so we have to use our resources as efficiently as possible to meet the users’ needs.”
What approach does the DOMINO-E project take to solve these kinds of problems?
“DOMINO-E introduces a federation layer that changes how requests are handled. Typically, users have to choose a specific system or satellite constellation to fulfill their requests. In DOMINO-E, users submit their requests to this federation layer, which then automatically dispatches them to the most suitable missions. This means users don’t need to worry about which system will handle their observation—it just gets done in the most efficient way possible.”
That sounds quite complicated. What makes it so challenging?
“One of the major challenges is the sheer number of possible decisions we have to evaluate. For example, if a task requires 60 images and there are two systems available, you end up with billions of billions of possible combinations. Our job is to find an efficient algorithm that can quickly sort through these possibilities and identify a good solution.”
And how do you approach this challenge in practice?
“We often start with an empty plan, then build on it iteratively, adding more observations step-by-step. This way, we can ensure that we’re maximizing the use of our available resources. Once we have an initial solution, we refine it by identifying and optimizing any weaker parts of the plan. This process, known as ‘any-time algorithms,’ allows us to continually improve the solution as long as we have time to compute.”
You let the algorithm run to come up with multiple solutions?
“Yes, exactly. We provide the algorithm with a set of rules and let it explore different solutions using some randomness in its approach. This ensures that we don’t end up with the same solution every time. During the search, the algorithm might generate thousands of potential solutions, and we can then choose the best one to present to the user.”
Can you describe a specific scenario where the DOMINO-E solution would significantly improve Earth observation capabilities?
“Currently, if you have a coverage request, you might need to rely on a single mission, which could be slow or limited in scope. With DOMINO-E’s multi-mission approach, the workload can be distributed across several systems, reducing the total time needed for coverage. This kind of efficiency is a game-changer in terms of how quickly and accurately we can deliver data to users.”
What’s the major challenge you face in developing the Coverage Service Domino?
“One of the main challenges is managing competition between different requests, all vying for the same resources. Users don’t always have visibility into the workload of each mission, and weather conditions can add another layer of complexity. A cloud cover, for instance, might force us to retake images. With a multi-mission setup like DOMINO-E, we can be more flexible, choosing the best satellite path to minimize the impact of such uncertainties.”
Is your focus more on making mission design more efficient or on handling image requests more effectively?
“Our main focus is on optimizing dispatching decisions. We don’t modify the missions or requests themselves but aim to distribute tasks in the most efficient way possible, taking into account factors like sensor size and the specific requirements of each request.”
What fascinates you most about working on these challenges?
“From a technical point of view, what’s fascinating is that solving these problems is like playing a complex game. We have an enormous number of potential solutions, and our task is to find a high-quality one. It’s both challenging and rewarding to develop algorithms that can navigate this vast decision space as efficiently as possible. From the point of view of the application, working on coverage requests for Earth observation is a challenge that contributes to key societal needs, such as providing to scientists and decision makers new data showing the impacts of the climate change over large regions.”
Thank you!