A data-driven parallel adaptive large neighborhood search algorithm for a large-scale inter-satellite link scheduling problem
Introduction
In recent years, aerospace technology and wireless communication technology have been rapidly developed, especially low-orbit communication satellite constellations, which have become increasingly important platforms in the satellite Internet of Things, remote sensing and telemetry, navigation enhancement and other fields [1]. For example, the OneWeb [2] and Starlink program [3] mainly serves global communication missions, including mobile services, fixed services, emergency communications, and new services such as Internet, Internet of Things, and multimedia. Simultaneously, with the continuous expansion of constellation scales and the introduction of new concepts such as "satellite-earth integration", " Space-Air integrated network" and "smart constellations", inter-satellite link (ISL) scheduling is already a very important technique [4] to solve inter-satellite information transmission and satellite-ground communication problems.
The ISL refers to a wireless communication link connecting two satellites, and the ISL scheduling means that each satellite in a constellation determines a connection scheme at different times. Meanwhile, the function of ISL is to transmit the satellite data outside a “measurement and control arc” back to the control department on the ground. By analyzing the ISL scheduling, three main characteristics are summarized below. (1) The low-orbit satellites are changing every moment with respect to the earth, which leads to a constant change in the whole constellation network. Therefore, the first characteristic of ISL scheduling is that the visible relationship between satellites changes continuously throughout the cycle. (2) Superabundant connection schemes between satellites are the second characteristic of ISL scheduling. For example, there are satellites in a constellation, the time cycle is , the average visible time is seconds between satellites, and a data transmission process takes seconds (including connection, transmission, and disconnection). Then, the number of schemes is . (3) Current satellite links use narrow-beam links, which allow satellites to establish a link only once within a certain time, and the satellite-to-satellite links are built if they are visible to each other. In addition, satellite link building needs to consider the orbit issue. Therefore, the complex constraint of building links between satellites is the third characteristic.
Compared with traditional ISL scheduling, the large-scale low-orbit communication satellite constellation inter-satellite link scheduling problem (LLCSC-ISLSP) studied in this paper is more complicated. It not only has ISL characteristics but also has the following difficulties: (1) Large constellation size. In general, the size of low-orbit communication satellite constellations is several to several dozen times larger than traditional navigation constellations. (2) The visible time between satellites is very short. Since the typical orbital altitude of low-orbit satellites is 500 km-1500 km, the system operation period is less than 2 h [5]. Therefore, the visible time is extremely short compared to the visible time of medium and high orbit satellites. (3) Millisecond data transmission speed. Although a low-orbit satellite has the advantages of small signal delay and low fading rate, it requires high data transmission speed, which needs to reach millisecond transmission time.
To address the LLCSC-ISLSP, this paper develops a time-discrete network multi-commodity flow (TDN-MCF) model. Based on this model, a data-driven parallel adaptive large neighborhood search (DP-ALNS) algorithm, an extension of the ALNS [34], is proposed for solving the problem. The design motivation of the model and algorithm is specified in Section 2.
The contributions of this paper are summarized as follows:
- 1
Large-scale low-orbit communication satellite constellation inter-satellite link scheduling problem. The ISL scheduling problem is initially solved in the navigation satellite system and small-scale low orbit satellite system, but the LLCSC-ISLSP has not been solved in previous research. Based on the ISL scheduling problem, the LLCSC-ISLSP is further developed in this paper and a comprehensive study of this problem is carried out. These works are beneficial for further research by other researchers studying ISL scheduling
- 2
Time-discrete network multi-commodity flow model. A TDN-MCF model is designed by combining the LLCSC-ISLSP and a multi-commodity flow model. First, the virtual topology strategy [17] is adopted to transform the continuous ISL scheduling problem into a time-discrete network. The satellite connection schemes of all timeslots in a cycle are expanded to build a network topology, where each satellite in each timeslot is viewed as a network node, entry satellites are viewed as nodes that can generate data, gateway satellites are regarded as transit nodes, and the gateway is viewed as the endpoint. In addition, virtual nodes are set up to ensure the integrity of the network. Finally, a multi-input, multi-output TDN-MCF model is obtained.
- 3
Data-driven parallel adaptive large-neighborhood search algorithm. The proposed DP-ALNS algorithm has the advantages of strong adaptability, large neighborhood search space and simple framework of the ALNS algorithm. Also, a probabilistic prediction model is trained in this algorithm by extreme gradient boosting (XGBoost) to predict the probability that a satellite is connected to its visible satellites. After obtaining the probabilistic prediction model, an initial solution generation strategy is used to generate high-quality initial solutions. Meanwhile, an adaptive mechanism consisting of a rule adaptive layer and an operator adaptive layer is designed, where the rule adaptive layer ensures that the algorithm can select appropriate rules and the operator adaptive layer improves the probability of searching for a better neighborhood solution during the search process. In addition, the parallel search mechanism and the dynamic programming idea are used to reduce the time consumption of the proposed algorithm.
The remainder of this paper is organized as follows. Section 2 presents the literature review and the motivation behind the proposed TDN-MCF model and the DP-ALNS algorithm. Section 3 describes the LLCSC-ISLSP and proposes the TDN-MCF model. Section 4 introduces the implementation of the DP-ALNS algorithm. Computational experiments using the proposed algorithm and detailed analysis are presented in Section 5, and the discussions and conclusions are presented in Section 6 and Section 7, respectively.
Section snippets
Literature review
At present, the main research objects about ISL scheduling are navigation satellite systems and small-scale low-orbit communication satellite constellations. To the best of our knowledge, few researchers have conducted in-depth research on the LLCSC-ISLSP. This section reviews previous studies based on the navigation satellite system inter-satellite link scheduling problem (NSS-ISLSP) and the small-scale low-orbit communication satellite constellation inter-satellite link scheduling problem
Problem description
The LLCSC-ISLSP is a multi-input multi-output problem, the objective function is defined as the shortest total time delay, and the network structure is composed of satellites and links. Meanwhile, the MCF problem is also a multi-input multi-output problem, the objective function is the shortest commodity transmission path, and the network is composed of nodes and edges. Therefore, a TDN-MCF model is proposed by combining the LLCSC-ISLSP and MCF in this paper. This is also the first study to
Data-driven parallel adaptive large neighborhood search algorithm
The framework of the DP-ALNS is built on the ALNS algorithm in Fig. 3. The ALNS has an ability to solve problems with many constraints and a large solution space and has the characteristics of simple structure and strong adaptation, where the core is that it can adaptively adjust the scores and selection weights of operators when solving different problems. Specifically, when operator selection process is carried out in each iteration, high-quality operators will be selected with a greater
Computational experiments
Twenty experiments are carried out to verity the effectiveness of the DP-ALNS algorithm. The results of the experiments are discussed in this section.
Discussion
The LLCSC-ISLSP is one of the issues that constellation management and control urgently needs to solve. If the problem can be solved well, it will obtain great military, commercial and civilian value. However, existing data-driven methods, metaheuristic approaches, and parallel scheduling have been used to solve lots of combinatorial optimization problems, but few algorithms have been proposed to solve combinatorial optimization problems similar to LLCSC-ISLSP characteristics.
Based on the ALNS
Conclusion
The study proposes a time-discrete network multi-commodity flow (TDN-MCF) model and a data-driven parallel adaptive large neighborhood search (DP-ALNS) algorithm to solve the inter-satellite link scheduling problem of large-scale low-orbit communication satellite constellations. Taking 900 satellite constellations as research objects, the average time delay is reduced by approximately 40% without affecting the communication quality.
The TDN-MCF model turns an ISL scheduling problem with
Author statement
Jinming Liu performed the model construction, algorithm design and experimental analysis, and wrote the manuscript;
Lining Xing contributed to the conception of the study and guided the research process;
Ling Wang helped guide the content and framework of the study;
Yonghao Du provided research ideas, helped guide the algorithm design and wrote the manuscript;
Jungang Yan helped perform problem model building and writing corrections;
Yingguo Chen helped perform the experimental design and the
Declaration of Competing Interest
The authors declared that they have no conflicts of interest to this work.
Acknowledgments
This study is supported by the National Natural Science Foundation of China (61773120, 72001212), the Special Projects in Key Fields of Universities in Guangdong (2021ZDZX1019), and the Hunan Postgraduate Research Innovation Project (CX20210083). Thanks are due to the referees for their valuable comments.
References (50)
- et al.
Interactions of the space debris environment with mega constellations—Using the example of the OneWeb constellation
Acta Astronaut
(2017) - et al.
Topological dynamics characterization for LEO satellite networks
Comput. Networks
(2007) - et al.
A scheduling strategy to inter-satellite links assignment in GNSS
Adv. Space Res.
(2021) - et al.
Optimization design of inter-satellite link (ISL) assignment parameters in GNSS based on genetic algorithm
Adv. Space Res.
(2017) - et al.
TDRSS scheduling algorithm for non-uniform time-space distributed missions
- et al.
A weighted semi-distributed routing algorithm for LEO satellite networks
J. Netw. Comput. Appl.
(2015) - et al.
Automating air force satellite control network (AFSCN) scheduling
Math. Comput. Model
(1996) - et al.
A Lagrangian heuristic for satellite range scheduling with resource constraints
Comput. Oper. Res.
(2011) - et al.
Optimization-based scheduling for the single-satellite, multi-ground station communication problem
Comput. Oper. Res.
(2015) - et al.
The time constrained multi-commodity network flow problem and its application to liner shipping network design
Transport. Res. Part E
(2015)