en fr Flexible Scheduling for Agile Earth Observing Satellites Production de plans flexibles pour des satellites agiles d’observation de la Terre Report as inadecuate

en fr Flexible Scheduling for Agile Earth Observing Satellites Production de plans flexibles pour des satellites agiles d’observation de la Terre - Download this document for free, or read online. Document in PDF available to download.

1 Toulouse - ONERA - The French Aerospace Lab

Abstract : Earth-observation satellites are space sensors which acquire data, compress and record it on board, and then download it to the ground. They evolve in a dynamic environment which implies several uncertain parameters such as cloud cover or available energy on board. These uncertainties make planning and scheduling satellite activities offline on the ground more and more arguable. Until now, plans are often built on the ground and are not to be modified on board. It means that in face of uncertainties, current plans need to be robust. It implies that on one hand, all the decisions are made offline on the ground and the satellite is a simple executive which neither makes, nor changes any decision and on the other hand, worst-case assumptions are made about uncertain parameters. is makes planning very pessimistic and plans suboptimal. This dissertation details our efforts at designing a exible decision-making scheme that allows to profit from the realization of uncertain parameters on board while keeping a fair level of predictability on the ground. Our first contribution concerns the data download problem. Sophisticated onboard algorithms compress low-interest zones of acquired images such as cloudy zones. The amount of data resulting from an acquisition is then unpredictable on the ground. Until now, all decisions regarding downloads are made on the ground, and data volumes are assumed to be maximum to make the plan robust. is makes plans highly suboptimal because real volumes are almost always lower than maximum and download windows are then underused. A flexible decision-making mechanism has been designed where only high-priority acquisitions are scheduled with worst-case assumptions. Other acquisition downloads are scheduled with expected volumes and conditioned by resource availability. The plan is then adapted on board: if acquisitions have a lower volume than predicted, acquisition downloads are added to the plan. If a low-priority acquisition has a higher volume than predicted, a look-ahead is performed to prevent future high-priority downloads to be endangered. This decision-making mechanism has been evaluated against pure ground and pure onboard approaches and showed great properties, ensuring predictability for the high-priority acquisitions and efficiency for the low-priority acquisitions while being computationally light on board. Our second contribution concerns the acquisition planning problem. When an acquisition plan is computed on the ground, resource availability such as memory or energy must be checked to ensure that the execution of the plan will not in endanger the satellite. From a technical point of view, because simulating the evolution of onboard energy and temperature is computationally expensive, it cannot be done finely during the planning process. It is replaced by high-level constraints with safety margins. The negative side of such a strategy is that the real energy profile is always better than predicted, and a lot of acquisitions that could have been done are eliminated when planning because of these high-level constraints. In a new decision-making scheme, these high-level constraints are removed for low-priority acquisitions. Observation plans produced on the ground are conditional plans involving conditions for triggering low-priority acquisitions. Compared with the current approach, this approach avoids wastage of resource and allows more acquisitions to be executed.


Author: Adrien Maillard -

Source: https://hal.archives-ouvertes.fr/


Related documents