Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Scheduling multiple agile Earth observation satellites with multiple observations

Xinwei Wang, Chao Han, Roel Leus

TL;DR

This paper addresses scheduling multiple agile Earth observation satellites (AEOS) with multiple observations per target, formulating it as a MAS with nonlinear profits. A column-generation framework based on Dantzig–Wolfe decomposition and a label-setting pricing solver is developed to solve the LP relaxation and generate effective columns, with integrality recovered by solving a final IP. Computational experiments on Gaojing-1 constellation data show an average optimality gap under 3% for AEOS instances, while the framework also offers strong performance for conventional CEOS scheduling. The work delivers a scalable, high-quality scheduling approach that accommodates multiple observations, sequence-dependent transformations, and resource constraints, with clear pathways for incorporating additional real-world factors in future work.

Abstract

The Earth observation satellites (EOSs) are specially designed to collect images according to user requirements. The agile EOSs (AEOS), with stronger attitude maneuverability, greatly improve the observation capability, while increasing the complexity in scheduling. We address a multiple AEOSs scheduling with multiple observations for the first time}, where the objective function aims to maximize the entire observation profit over a fixed horizon. The profit attained by multiple observations for each target is nonlinear to the number of observations. We model the multiple AEOSs scheduling as a specific interval scheduling problem with each satellite orbit respected as machine. Then A column generation based framework is developed to solve this problem, in which we deal with the pricing problems with a label-setting algorithm. Extensive simulations are conducted on the basis of a China's AEOS constellation, and the results indicate the optimality gap is less than 3% on average, which validates the performance of the scheduling solution obtained by the proposed framework. We also compare the framework in the conventional EOS scheduling.

Scheduling multiple agile Earth observation satellites with multiple observations

TL;DR

This paper addresses scheduling multiple agile Earth observation satellites (AEOS) with multiple observations per target, formulating it as a MAS with nonlinear profits. A column-generation framework based on Dantzig–Wolfe decomposition and a label-setting pricing solver is developed to solve the LP relaxation and generate effective columns, with integrality recovered by solving a final IP. Computational experiments on Gaojing-1 constellation data show an average optimality gap under 3% for AEOS instances, while the framework also offers strong performance for conventional CEOS scheduling. The work delivers a scalable, high-quality scheduling approach that accommodates multiple observations, sequence-dependent transformations, and resource constraints, with clear pathways for incorporating additional real-world factors in future work.

Abstract

The Earth observation satellites (EOSs) are specially designed to collect images according to user requirements. The agile EOSs (AEOS), with stronger attitude maneuverability, greatly improve the observation capability, while increasing the complexity in scheduling. We address a multiple AEOSs scheduling with multiple observations for the first time}, where the objective function aims to maximize the entire observation profit over a fixed horizon. The profit attained by multiple observations for each target is nonlinear to the number of observations. We model the multiple AEOSs scheduling as a specific interval scheduling problem with each satellite orbit respected as machine. Then A column generation based framework is developed to solve this problem, in which we deal with the pricing problems with a label-setting algorithm. Extensive simulations are conducted on the basis of a China's AEOS constellation, and the results indicate the optimality gap is less than 3% on average, which validates the performance of the scheduling solution obtained by the proposed framework. We also compare the framework in the conventional EOS scheduling.

Paper Structure

This paper contains 26 sections, 5 equations, 4 figures, 11 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Literature review
  3. EOS scheduling
  4. AEOS scheduling
  5. Exact methods.
  6. Heuristics and metaheuristics.
  7. Learning-based methods.
  8. Discussion.
  9. Definitions and problem statement
  10. MAS
  11. Extensions
  12. Column generation
  13. Dantzig-Wolfe decomposition
  14. The pricing problems
  15. Column initialization
  16. ...and 11 more sections

Figures (4)

  • Figure 1: Illustration of the fixed observation interval of a CEOS. The CEOS can only maneuver along the roll axis, so that its OTW coincides exactly with the VTW.
  • Figure 2: Comparison of the observation capability of CEOS and AEOS. Unlike a CEOS, which can only observe within its VTW, an AEOS has additional pitch-axis maneuverability that allows it to look ahead or look back. This enables multiple potential OTWs within the same VTW, thereby increasing flexibility and observation opportunities.
  • Figure 3: Overview of the MAS modeling process.
  • Figure 4: Illustration of the profit function for a single target $i$ with up to four observations ($N_i=4$). The profit values ($\pi_{i0}=0$, $\pi_{i1}=1$, $\pi_{i2}=3$, $\pi_{i3}=6$, $\pi_{i4}=10$) are illustrative examples to demonstrate the nonlinear increase.

Theorems & Definitions (4)

  • Definition 1
  • Definition 2
  • Definition 3
  • proof