1. Introduction
Remote sensing satellites are particularly effective tools for observing and analyzing earth’s resources and environment and play an irreplaceable role in earth resource exploration, natural disaster prevention, and environmental protection [
1]. Compared to traditional non-agile remote sensing satellites, as shown in
Figure 1, the advantage of agile remote sensing satellites lies in their flexible and fast maneuverability, which not only greatly enhances the observation capabilities of satellites but also leads to a significant increase in the complexity of solving satellite execution plans [
2].
Before the rapid development of agile remote sensing satellites, many scholars had conducted research over a long period on the scheduling problem of non-agile remote sensing satellites. Previous research mainly focused on explaining the concept of satellite scheduling. Due to the lack of large-scale task sets, satellite scheduling models were often simplified [
3,
4,
5,
6]. Abramson [
7] considered satellite constraints, including satellite lateral sway and storage and introducing an integer linear programming method for finding a solution. Bianchesi [
8] takes the COSMO SkyMed constellation as the research object, with the goal of maximizing the number of images obtained, and adopts a heuristic construction algorithm to solve the satellite scheduling problem, but does not introduce enough real constraints. Florio [
9] considered the mobility constraints of satellites in his research, while also considering the energy and storage capabilities of satellites as constraints. Lemaître [
10] took into account the strong maneuverability of agile remote sensing satellites and considered the satellite’s maneuverability model in detail, but simplified the processing by incorporating additional constraints for the satellite. Reference [
11] treated the decomposition of regional targets as a type of constraint for satellites. Reference [
12] proposed and to some extent solved the resource matching problem based on conventional observation tasks and emergency observation tasks, with task efficiency as the global optimization objective in the scheduling process. Peng’s research also focuses on this aspect [
13]. Xu [
14] focused on satellite observation time window constraints and energy constraints, simplifying other constraints. Reference [
15] focused on addressing the observation time dependency in the scheduling process of agile remote sensing satellites in his research, but simplified other constraints. The algorithm proposed by Reference [
16] can effectively improve the real-time performance of computation, but its feasibility for large-scale tasks has not been verified. Du’s research focuses on multi-star mission scheduling, and his proposed method is beneficial for the efficiency of large-scale task allocation, but does not consider the single-star constraint problem in detail [
17].
Starting from the end of the twentieth century, intelligent heuristic optimization algorithms have been widely used in the field of satellite mission planning and scheduling due to their good optimization capabilities, such as the ant colony algorithm [
18,
19], the particle swarm optimization algorithm [
20], the genetic algorithm [
21,
22], the improved genetic algorithm [
23,
24,
25,
26,
27], the simulated annealing algorithm [
28,
29,
30], the taboo search algorithm [
31,
32,
33,
34,
35], and evolutionary algorithms [
36,
37]. The complexity of agile remote sensing satellite mission scheduling based on large-scale tasks has increased exponentially, and intelligent heuristic optimization algorithms have shown the disadvantage of low solution efficiency during the solution process. Therefore, traditional algorithms are in urgent need of further innovation and development. In recent years, artificial intelligence has been applied in various fields. Many scholars have combined artificial intelligence algorithms with satellite scheduling problems and conducted a series of studies [
38,
39,
40,
41,
42,
43]. In addition, many scholars have applied physics-informed deep learning approaches to many fields and proved the algorithm’s superior optimization ability [
44]. Similarly, it is feasible to apply this method to the field of satellite mission scheduling. The biggest feature of this type of method is that it can improve the timeliness of satellite execution plans while achieving global optimization. However, these methods are limited by the performance of on-board computers and have certain limitations in the current field of engineering applications.
Currently, in the engineering practice process of satellite mission scheduling, how to efficiently and reliably generate satellite task execution plans is a key issue that urgently needs to be solved. Narayanan [
45] integrated the ideas of genetic algorithms and quantum theory and demonstrated that this algorithm has a significant effect on improving search capability and computational speed. Silveira [
46] innovatively proposed a quantum-inspired evolutionary algorithm and demonstrated its effectiveness through sorting problems. Nowotniak [
47] introduced the concept of high-order on the basis of a quantum genetic algorithm and proved that this method has significant effects on improving computational speed.
In our study, we propose a novel MAS-HOQGA to solve the agile remote sensing satellite scheduling problem. First, taking into account the constraints, including the time-dependent characteristics of agile remote sensing satellites, attitude maneuverability, payload observation capability, data transmission resources, storage capability, and energy balance, a refined satellite constraint model with comprehensive multiple constraints was established. The scheduling objective function was established based on the comprehensive value derived from the number of tasks and their priorities. Then, on the basis of the traditional QGA, the quantum register operator, the adaptive evolution strategy, and the adaptive mutation transfer strategy are introduced. The proposed MAS-HOQGA is used to solve the agile satellite scheduling problem.
Compared to traditional QGA, the quantum chromosome in this paper is composed of quantum register operators, ensuring that the proposed algorithm has advantages in both individual measurement and updates. Quantum registers can reduce the time complexity of individual measurements and updates, thereby improving algorithm performance. In addition, the adaptive evolution strategy and adaptive mutation transfer strategy introduced in this paper are beneficial for improving the convergence speed of the algorithm and preventing it from falling into local optima.
The experimental results show that the MAS-HOQGA has achieved significant improvements in both comprehensive revenue and algorithm running time for scheduling results compared with the QGA.
Figure 2 shows the framework of the research content of this paper.
Specifically, the specific work and innovation of the paper are as follows:
(1) Considered the agile remote sensing satellite mission scheduling problem for large-scale tasks scenarios.
(2) Established a comprehensive multi-constraint satellite scheduling model and described the optimization objective function based on the total task number and the total task priorities.
(3) A higher-order QGA based on multi-adaptive strategies was proposed.
(4) The proposed algorithm shows excellent computing power and global optimization capabilities in the field of agile remote sensing satellite mission scheduling.
This paper is divided into five parts. After the introduction,
Section 2 introduces the establishment process of the multi-constraint refined satellite scheduling model.
Section 3 introduces MAS-HOQGA, including quantum individual encoding and quantum register initialization, the measurement and adaptive evolution strategies of quantum register operators, and the adaptive mutation transfer strategy of quantum registers. In
Section 4, the comprehensive revenue and algorithm running time of the MAS-HOQGA and QGA for scheduling tasks of different scales are compared, and the impact of the probability amplitude adjustment parameter setting on the results is analyzed. The final section presents the conclusions of this paper.
2. Refined Satellite Scheduling Model with Multi-Constraints
The establishment of a satellite mission scheduling model needs to consider the usage constraints of the satellite and the requirements of its various subsystems. In this section, the relevant parameters in the satellite scheduling model are defined, and the constraints of each subsystem in the satellite scheduling model are given.
2.1. Parameter Definition
The satellite properties are defined as follows:
: satellite storage maximum;
: satellite payload data rate;
: satellite energy maximum;
: the energy consumption per unit of time during satellite payload operation;
: the energy consumption per unit of time during satellite maneuvering;
: the angle at which a satellite can maneuver its orientation per unit of time;
Fov: the satellite payload field of view angle.
The task properties are defined as follows:
: the collection of tasks, where N is the total number included in , and is the i-th observation task.
: is the task ID, is the task priority, is the task latitude, is the task longitude, is the task altitude, is the task observation duration, is the task observation began time, and is the task observation end time.
The task window properties are defined as follows:
: the observation windows set of task , m is the total number of observation windows corresponding to the mission , where , is the number of cycles in which the observation window is located, is the beginning time of the window, and is the end time of the window.
2.2. Scheduling Model Constraints
2.2.1. Payload Constraints
Based on the working capacity of the payload, specify the maximum working time within one orbit of the payload and require that the working time of each orbit of the payload be less than the maximum working time.
where
k is the number of observation tasks arranged within the
j-th orbital cycle,
is the observation duration of the
i-th payload mission arranged within the orbital cycle
j, and
is the maximum working duration of the payload within one orbital cycle.
2.2.2. Maneuverability Constraints
The time interval between two observation tasks needs to be greater than the fastest maneuvering time of the satellite, as shown in
Figure 3.
2.2.3. Energy Constraints
During the operation of the satellite, it is necessary to always ensure the satellite energy balance; that is, the energy consumed by the satellite during operation should be less than the maximum energy consumption.
N is the count of successfully arranged observation tasks,
N + 1 is the number of attitude maneuvers arranged,
is the initial attitude of the satellite, and
is the attitude that the satellite needs to maintain after completing all planned tasks.
2.2.4. Data Storage Constraints
According to the design requirements for satellite data storage, it is required that the data generated by the payload be promptly arranged to the ground, and they cannot exceed the satellite storage maximum.
2.2.5. Comprehensive Revenue Scheduling Function
In the actual engineering scheduling process, satellite users hope that the satellite can observe more tasks on one hand, and on the other hand, they also hope that the satellite can prioritize completing higher priority tasks. This paper considers both the number of observation tasks and their corresponding priorities. A comprehensive scheduling revenue function is established as the objective function for the scheduling solution.
where
represents the revenue weight corresponding to the number of observation missions,
represents the revenue weight corresponding to the priority of the observation missions, and
serves as a decision variable.