Genetic algorithm–based scheduling for ground ...

Paper Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016) DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89

Genetic algorithm–based scheduling for ground support of multiple satellites and antennae considering operation modes Junghyun Lee* Gwangju Institute of Science and Technology (GIST), Gwangju 61005, Republic of Korea

Haedong Kim** Korea Aerospace Research Institute (KARI), Daejeon 34133, Republic of Korea

Hyun Chung*** Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, Republic of Korea

Kwanghee Ko**** Gwangju Institute of Science and Technology (GIST), Gwangju 61005, Republic of Korea

Abstract Given the unpredictability of the space environment, satellite communications are manually performed by exchanging telecommands and telemetry. Ground support for orbiting satellites is given only during limited periods of ground antenna visibility, which can result in conflicts when multiple satellites are present. This problem can be regarded as a scheduling problem of allocating antenna support (task) to limited visibility (resource). To mitigate unforeseen errors and costs associated with manual scheduling and mission planning, we propose a novel method based on a genetic algorithm to solve the ground support problem of multiple satellites and antennae with visibility conflicts. Numerous scheduling parameters, including user priority, emergency, profit, contact interval, support time, remaining resource, are considered to provide maximum benefit to users and real applications. The modeling and formulae are developed in accordance with the characteristics of satellite communication. To validate the proposed algorithm, 20 satellites and 3 ground antennae in the Korean peninsula are assumed and modeled using the satellite tool kit (STK). The proposed algorithm is applied to two operation modes: (i) telemetry, tracking, and command and (ii) payload. The results of the present study show near-optimal scheduling in both operation modes and demonstrate the applicability of the proposed algorithm to actual mission control systems. Key words: genetic algorithm, mission control system, ground support, visibility conflict, scheduling optimization, operation mode

1. Introduction

which are transmitted and received within the radio visibility between ground antennae and satellites. Satellites are largely operated by manual commands owing to the unpredictability of the space environment. Satellite operational missions are largely divided into two modes: 1) telemetry, tracking, and command (TTC) for maintaining orbit, tracking satellites, and the state of health (SOH); 2) payload (PL) for undertaking the

Communication between a satellite and a ground antenna is performed when the satellite passes through the ground antenna’s “visibility”—i.e., the cone-shaped region of communication extending from the antenna. Satellites are controlled and operated using tele-commands and telemetry,

* PhD Student Professor ** *** Professor **** Professor, Corresponding author: [email protected]

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/bync/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Received: July 27, 2015 Revised: November 24, 2015 Accepted: February 18, 2016 Copyright ⓒ The Korean Society for Aeronautical & Space Sciences

(89~100)15-125.indd 89

89

http://ijass.org pISSN: 2093-274x eISSN: 2093-2480

2016-03-29 오후 7:27:32

Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016)

the conflict problem of satellite visibility via a deterministic approach, which is impractical. The Indian Space Research Organization (ISRO) devised a model using a genetic algorithm to solve conflicts between Indian Remote Sensing satellites (IRSs) [1, 24, 25]. Castaing [26] established visibility scheduling for multiple satellites and multiple ground control centers. ISRO and Castaing claimed that they used a genetic algorithm to solve the visibility conflict problem but did not present any detailed information about the methodologies, which means that their approach could not be reconstructed. Gwangju Institute of Science and Technology (GIST) and KARI presented a solution to the conflict problem of visibility using multi-antenna scheduling (MAS) [27] and a genetic algorithm [28]. Corrao et al. [29] proposed a resolution of the visibility conflict between multiple satellites and a single ground antenna by using a wide spectrum of algorithms. However, their research did not consider antenna support time and real operation modes. Xhafa et al. [30] and Spangelo et al. [31] studied the allocation of communication data within the visibility between multiple satellites or ground antennae and a single ground antenna or; a single satellite and multiple ground antennae, respectively. However, they did not consider multiple satellites and multiple ground antennae. Xhafa et al. [32, 33] developed an algorithm for numerous satellites and ground antennae using a genetic algorithm; however, operational factors were not acquired and verification was not performed for various operating modes.

mission of the satellite and sending the acquired data via a downlink [1]. When two or more satellites simultaneously pass over a ground station, visibility conflicts occur because most ground antennae can only support one satellite at a time. As the number of satellites increases, the frequency of visibility conflicts increases; additionally, allocation of communication data within the visibility becomes problematic, and operational and mechanical considerations increase exponentially. Currently, mission control operations, including satellite–antenna scheduling, are performed manually [2-5]. However, manual scheduling generates unexpected errors and is inefficient. The mission control system for the Korea Multi-Purpose Satellite (KOMPSAT) series was developed by the Korea Electronics and Telecommunications Research Institute (ETRI) and Korea Aerospace Research Institute (KARI) [68]. KOMPSATs are low Earth orbit (LEO) satellites following a sun-synchronous orbit of Korea; their main mission is acquiring images of Earth. Currently, KARI operates KOMPSAT 2, 3, 3A, and 5 with a very low frequency of visibility conflicts, because their orbital parameters are designed to avoid visibility conflicts. However, additional KOMPSATs will soon be introduced [9]. Hence, an algorithm is needed that can automatically solve ground support problems due to visibility conflicts.

1.1 Previous work Scheduling optimization problems have mainly been studied in industrial engineering [10, 11]. Specifically, the ground support allocation problem is similar to assigning N jobs to M machines or a multiple-knapsack problem; the difference, however, lies in the fact that the visibility (resource) of one satellite can block that of others, and it exists for only a predetermined period before expiring [1, 12, 13]. Many studies have been performed on scheduling methods for satellite missions. KARI developed an automated algorithm for mission scheduling that focuses on a single satellite [14-16]. Dishan et al. [17] and Frank et al. [18] proposed dynamic and static scheduling methods for Earth observation (EO) satellites. However, they proposed heuristic methods for scheduling that do not guarantee the optimality of their solution. Lin et al. [19, 20] studied the daily image scheduling of a low-orbit EO satellite and solved a scheduling problem for single satellite operation. The methodology of visibility conflicts has also been studied. Globus et al. [21, 22] applied various stochastic algorithms and made comparative analysis of the mission operation of EO satellites. ArKari et al. [23] tried to solve

DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89

(89~100)15-125.indd 90

1.2 Proposed model In this study, we resolve the ground support problem during periods of visibility via a genetic algorithm for the KOMPSAT series. The algorithm provides maximum benefits to the user by taking into account various considerations in terms aspects of actual operation such as user priority, operational urgency, profit, contact interval, support time, satellite performance, and benefit. For the simulation, we assume multiple satellites with sun-synchronous orbits and multiple ground antennae with visibility conflicts, which are modeled using STK. The proposed algorithm is verified for the TTC and PL modes. We demonstrate that this algorithm can be applied to real operations. The contributions of this paper are as follows. 1) We propose an algorithm for antenna support scheduling of multiple satellites and ground antennae. Unlike the majority of previous related papers, which merely considered mathematical modeling of scheduling, this paper has practical value because the proposed method can be applied to the real operations of the KOMPSAT series. 2) We

90

2016-03-29 오후 7:27:32

Junghyun Lee Genetic algorithm–based scheduling for ground support of multiple satellites and antennae considering .....

design an optimization formula and a genetic algorithm for 2.2 Visibility conflict solving such a problem. Generally, the satellite-scheduling When two or more satellites simultaneously pass over a problem is regarded as an NP-hard problem [34-38]. In this ground antenna or the AOS of a following satellite occurs case, a deterministic formula cannot be applied, and every before the LOS of the prior satellite, a visibility conflict possible scheduling should be searched. Therefore, we occurs—i.e., each satellite blocks the visibility of the other creatively design decision variables and a chromosome to satellite—as illustrated in Fig. 2. The visibility to which apply a stochastic approach. 3) We account for scheduling communication data is allocated should be selected considerations and operation modes in the proposed cannot be applied, and every possible scheduling should be searched. Therefore, we creatively design according to the mission and operation mode. The difference algorithm such that a near-global optimal scheduling result between the last3)LOS and the AOS is designated as the decision variables and a parameters chromosome toand apply a stochastic approach. We account forfirst scheduling is obtained despite numerous scheduling total visibility time (TVT). mechanical constraints.considerations and operation modes in the proposed algorithm such that a near-global optimal The rest of this paper is organized as follows. In Section scheduling result is obtained despite numerous scheduling parameters and mechanical constraints. 2.3 Support and reconfiguration of antenna 2, we introduce concepts related to visibility and associated The rest of this paper is organized as follows. In Section 2, we introduce concepts related to conflicts. Section 3 defines the scheduling parameters. The Actual communication is executed through the support modeling and formulation areand presented Section 4. In3 defines the scheduling parameters. The modeling and visibility associated in conflicts. Section of the ground antenna within the visibility. Specifically, the Section 5, a genetic algorithm is designed. Section 6 details theSection satellite moving along the orbit, formulation are presented in Section 4. In Section 5, ground a genetic antenna algorithm istracks designed. 6 details the simulation conditions for applying the algorithm. In and the line of sight of the antenna points to the satellite. the simulation conditions for applying thewe algorithm. In Section 7, results are presented and analyzed. Section 7, results are presented and analyzed. Finally, When the antenna support transitions from one satellite summarize and conclude our study in Section 8. Finally, we summarize and conclude our study in Section 8. to another, some reconfiguration time is required for mechanical movement or software initialization.

2. Background 2.1 Visibility

2. Background

3. Scheduling considerations

2.1 Visibility

Figure 1(a) illustrates the visibility of a satellite and ground antenna. The start and end times of

Satellites passing over ground antennae cannot Figure 1(a) illustrates the visibility of a satellite and visibility are defined as the acquisition of signal (AOS) and loss of signal (LOS), respectively. The communicate simultaneously in multiple-satellite ground antenna. The start and end times of visibility are actual communication is executed by allocating communication data within the visibility while operation. Consequently, satellites are selected for efficient defined as the acquisition of signal (AOS) and loss of signal (LOS), respectively. Theconsidering actual communication is executed the transmission rate, as shown Fig. 1(b).communication such that no supporting conflict arises. The following seven conditions are taken into account in by allocating communication data within the visibility while selecting satellites and ground antennae for communication. considering the transmission rate, as shown Fig. 1(b).

( a ) Visibility of satellite and ground antenna ( b ) Data allocation during visibility

(a) Visibility of satelliteFigure and ground antenna between satellite (b) Dataand allocation 1. Communication ground during visibility 2.2 Visibility Fig. 1. Communication between satelliteconflict and ground difference between themore last LOS and the first AOS is designated the total visibility time (TVT). When two or satellites simultaneously pass over aas ground antenna or the AOS of a following satellite occurs before the LOS of the prior satellite, a visibility conflict occurs—i.e., each satellite blocks the visibility of the other satellite—as illustrated in Fig. 2. The visibility to which communication data is allocated should be selected according to the mission and operation mode. The

(a) Multiple satellites and a single ground antenna (a) Multiple satellites and a single ground antenna

(b)

Total visibility time (b) Total visibility time

Figure 2. Visibility conflict of multiple satellites

Fig. 2. Visibility conflict of multiple satellites 2.3 Support and reconfiguration of antenna

Actual communication is executed through the support of the ground antenna within the visibility.

91 along the orbit, and the line of sight of the Specifically, the ground antenna tracks the satellite moving

http://ijass.org

antenna points to the satellite. When the antenna support transitions from one satellite to another, some reconfiguration time is required for mechanical movement or software initialization.

(89~100)15-125.indd 91

2016-03-29 오후 7:27:33

Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016)

3.1 User priority (W1)

support time is determined from the RST as

This parameter represents the priority that users determine arbitrarily. It represents the value of the requesting customer. Users can utilize this value to intervene in the scheduling algorithm in case of an unexpected situation.

3.2 Emergency (W2) This parameter is given in accordance with the level of urgency. The SOH of satellites is regularly monitored. Satellites with SOH-related urgency must have a larger weight than other satellites. A value of 100 is used in the control operation for at-risk satellites, and 0 is used for healthy satellites.

W5 =

1000 . 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅

(1)

3.63.6Satellite performance Satellite performance (W6(W ) 6)

This parameter In an imaging mission, a sa This parameterrepresents represents the the quality quality of of payload payload data. data. In an imaging mission, a satellite with a higher-precision lens higher-precision lens generatesimages; more sophisticated it is given a larger w generates more sophisticated therefore, itimages; is giventherefore, a larger weight. Table 2 summarizes the performance weights. 2 summarizes the performance weights.

3.7 Remaining resource (W7)Table 2. Weights of satellite performance Satellite performance

Weights

The remaining resources of a satellite are also considered in satellite selection. A satellite consumesHigh resources such45 Middle as electric power and onboard memory to obtain ground30 3.3 Profit (W3) Low images. Therefore, satellites with fewer remaining resources15 Profit expresses the financial benefits of contacting the 1000 must be selected preferentially (i.e., given larger weights) W5 = . (1) 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 a satellite. Among multiple satellites, a satellite that provides to downlink their stored data. Table 3 gives the weights for 3.7 Remaining resource (W7 ) higher profit is preferably selected. For an imaging mission, remaining resources. the profit is the price of the images acquired by the satellite. The remaining resources of a satellite are also considered in satellite selection. A satell 3.6 Satellite performance (W6 ) monitored. Satellites with SOH-related urgency must have a larger weight than other satellites. A resources suchdescription as electric power and onboard memory to obtain ground images. Theref 3.4 Contact interval (W4) Problem This parameter 4. represents the quality of payloadand data.formulation In an imaging mission, a satellite with a value of 100 is used in the control operation for at-risk satellites, and 0 is used for healthy satellites. with fewer remaining resources must be selected preferentially (i.e., given larger weights Larger weights are assigned to satellites higher-precision that have lens We generates more sophisticated images; therefore, is given a larger weight. Table assume a situation with numerous satellitesitand ground been contacted less recently. Given that satellite SOH is their stored data. Table 3 gives the weights for remaining resources. 1000 antennae with visibility conflicts. The ground antennae are 25summarizes W =contact . the performance weights. (1) periodically, if a satellite does not make ) 3.3 Profit (W3monitored 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 located close to each other, and each can satellite resources Table 3.support Weights one of remaining with the ground antenna for an extended period of time, its Table 2.via Weights of satellite performance Profit expresses the financial benefits of contacting the satellite. Among multiple satellites, at a atime. Support the ground antennae is required for SOH cannot be checked, which affects mission planning. Remaining resources Weight communication.Satellite The start and end Weights of the RST are called performance satellite that provides a higher if profit is preferably selected.does For an imaging the profit is the (W ) 3.6mission, Satellite performance For example, an imaging satellite not make contact 6 the support start time (SST) and support0–20% end time (SET),30 High 45 with acquired the ground price of the images by theantenna satellite. for a long time, the image data will 20–40% respectively. An example of RST allocation for onemission, ground24a satellite with a This parameter represents the quality of payload data. In an imaging Middle 30 remain on the onboard memory and the satellite cannot 40–60% antenna and three satellites is shown in Fig. 3. The RST of a18 execute its inherent mission. Hence, the satellite must Low 15 higher-precision lens generates more sophisticated therefore, it is given a larger weight. Table satellite is allocated within theimages; TVT. Reconfiguration time is12 60–80% be preferentially selected. Considerations for the contact 3.4 Contact interval (W4 ) required in order to support other satellites after allocation 80–100% 6 2 summarizes the performance weights. interval are presented in Table 1. 3.7 Remaining resource (W ) Larger weights are assigned to satellites that have been contacted less recently. Given that satellite 7

Table 2. Weights of satellite performance Table 2. Weights of satellite performance 4. Problem description andconsidered formulation SOH is monitored periodically, if a satellite does not make contact with the ground antenna for an 3.5 Support time (W5) The remaining resources of a satellite are also in satellite selection. A satellite consumes Satellite performance Weights assume with numerous satellites and ground antennae with visibility c extended period of time, its SOH cannot be checked, which affects mission planning.such For example, ifWepower resources as electric anda situation onboard The required communication time for each satellite, Highmemory to obtain 45 ground images. Therefore, satellites

i.e., the support (RST), should be athe image data an imaging satellite doesrequired not make contact withtime the ground antenna for a also long time, Middle 30 (i.e., antennae locatedpreferentially close to each other, and larger each can support satellite at a t with fewer remainingground resources must beareselected given weights) to one downlink

consideration in satellite selection. A satellite with a long Low 15 via3 the ground antennae is requiredresources. for communication. The start and end of the RST a RST brings less benefit to the user. Therefore, such satellites their stored data. Table gives the weights for remaining satellite must should be preferentially selected. Considerationsinforselection. the contact The interval are presented in Table be given less preference weight of support startTable time and endresources time (SET), respectively. An example of RST Table 3. Weights of remaining resources 3.(SST) Weights ofsupport remaining 3.7 Remaining resource (W 7)

will remain on the onboard memory and the satellite cannot execute its inherent mission. Hence, the

1.

Table 1. Weights of contact intervals Table 1. Weights of contact intervals Contact interval

Weight

1–2 hours

5

3–4 hours

10

5–6 hours

15

7+ hours

20

Remaining resources The remaining resources of a satellite are also consideredWeight in satellite selection. A satellite consumes 0–20%

30

20–40%

24

resources such as electric power and onboard memory to obtain ground images. Therefore, satellites 40–60%preferentially18(i.e., given larger weights) to downlink with fewer remaining resources must be selected 60–80%

12

80–100%

6

their stored data. Table 3 gives the weights for remaining resources.

3.5 Support time (W5 )

Table 3. Weights of remaining resources

The required communication time for each satellite, i.e., the required support time (RST), should and formulation 4. Problem description Remaining resources

Weight DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89 also be a consideration in satellite selection. A satellite with a long RST brings lessassume benefit92 the user. with numerous 0–20% 30 antennae with visibility conflicts. The We ato situation satellites and ground Therefore, such satellites should be given less preference in selection. The weight of support time is

20–40%

24

40–60%

18

ground antennae are located close to each other, and each can support one satellite at a time. Support

determined from the RST as

(89~100)15-125.indd 92

60–80% via the ground antennae is required for communication. The12start and end of the RST are called the 80–100% 6

support start time (SST) and support end time (SET), respectively. An example 2016-03-29 of RST allocation for 오후 7:27:33

satellite antenna. The RSTs satellites allocated corresponding antennae satellite andand antenna. The RSTs of of satellites areare allocated to to thethe corresponding antennae se

satellite antenna. RSTs satellites are As allocated to thecontinues, corresponding antennaeR fromthetheand first rowto The tothethe array. Asallocation allocation continues,visibilities, visibilities from first row lastlastof of of thethearray.

Junghyun Lee Genetic algorithm–based scheduling for groundsatellite support offirst multiple and antennae considering antenna. The RSTs of of satellites are allocated to.....the corresponding antennae from theand row satellites to the last the array. As allocation continues, visibilities

reconfigurations conflict with each other. a conflict occurs, communication is not reconfigurations conflict with each other. If aIfconflict occurs, thethe communication is not ex

the first row to with the last of the As allocation visibilities, reconfigurations conflict each other. Ifarray. a conflict occurs, the continues, communication is not the parameter corresponding the index is assigned zero. Otherwise, parameter c thefrom parameter corresponding to to the index is assigned as as zero. Otherwise, thethe parameter corr variable characterizes the order of the satellite index and conflict with other. If adesign conflict the communication is not the parameter corresponding toeach the index isThe assigned asoccurs, zero.ischaracterizes Otherwise, the to the executed index is given as one. The design variable characterizes the order the randomness of the parameter to reconfigurations the executed index setset isantenna given asindex. one. The variable theparameter order of oct determined according to iswhether the communication the parameter corresponding to the index isThe assigned asvariable zero.is the parameter to the executed index set as one. design the order index randomness the antenna index. The parameter ischaracterizes determined according t index andand thethe randomness of of thegiven antenna index. The parameter isOtherwise, determined according to ocw performed. A scheduling example is explained in Section to the and executed index set isof as one.example The design characterizes thewith order of index the israndomness the antenna index. The parameter is determined according to 5.1 with a chromosome. communication is performed. A scheduling example is variable explained in Section a ch Figure 3. Example of support time allocation communication performed. Agiven scheduling is explained in Section 5.15.1 with a chrom

to the previous satellite. Communication among multiple satellites and a ground antenna can be converted to a taskscheduling problem by treating the RST and visibility as task and resource, respectively, to be allocated within the TVT.

4.14.1Design variables parameters Design variables andand parameters

index and the randomness of the antenna index. The parameter is in determined according communication is performed. A scheduling example is explained Section 5.1 with a chto

4.2 Objective function

Design variables and is performed. Design variables andparameters parametersare areobtained obtainedconsidering considering a chromosome in a function genetic algorithm.A scheduling example is explained in Section 5.1 with a chr 4.2 Objective function 4.2communication Objective a chromosome in a genetic algorithm. The problem of The optimization problem has the following objective The problem of scheduling satellitesisand visibilities asoptimization an NP-hard problem. 4.2 Objective function scheduling multiple satellitesmultiple and visibilities regarded as is regarded The problem following objective function: function: The optimization problem hashas thethe following objective function: an NP-hard problem. Accordingly, the designed formula is 4.2The Objective function 𝑁𝑁𝑁𝑁 Accordingly, the designed formula is stochastic–deterministic. The overall variables are𝑁𝑁𝑁𝑁designed in optimization problem has the following objective function: stochastic–deterministic. The overall variables are designed � (3) � maximize 𝑓𝑓𝑓𝑓�𝑋𝑋𝑋𝑋 = � 𝑅𝑅𝑅𝑅 × 𝑝𝑝𝑝𝑝 � maximize 𝑓𝑓𝑓𝑓�𝑋𝑋𝑋𝑋� = � 𝑁𝑁𝑁𝑁𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 ×𝑖𝑖𝑖𝑖 𝑝𝑝𝑝𝑝𝑖𝑖𝑖𝑖 , 𝑖𝑖𝑖𝑖 , The optimization 𝑖𝑖𝑖𝑖=1 problem has the following objective function: in an an N forfor N satellites: N ××3 3array array N satellites: 𝑖𝑖𝑖𝑖=1 maximize 𝑓𝑓𝑓𝑓�𝑋𝑋𝑋𝑋�� = � 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 × 𝑝𝑝𝑝𝑝𝑖𝑖𝑖𝑖 , 𝑁𝑁𝑁𝑁 � (2) 𝑋𝑋𝑋𝑋 = [Satellite index𝑖𝑖𝑖𝑖 , Antenna index𝑖𝑖𝑖𝑖 , pi ], pi ∈ {0, 1} . (2) ∑7𝑗𝑗𝑗𝑗=1 =7𝑗𝑗𝑗𝑗=1 𝐹𝐹𝐹𝐹𝑗𝑗𝑗𝑗𝑖𝑖𝑖𝑖=1 , is𝑤𝑤𝑤𝑤 is the weight of scheduling, is the ratio depends where where weight of F𝐹𝐹𝐹𝐹is𝐹𝐹𝐹𝐹isthe 𝐹𝐹𝐹𝐹𝑗𝑗𝑗𝑗 𝑤𝑤𝑤𝑤 ,𝑤𝑤𝑤𝑤w isthe the weight ofscheduling, scheduling, the ratio thatthat depends on on the where 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 𝑅𝑅𝑅𝑅=𝑖𝑖𝑖𝑖 ∑ 𝑗𝑗𝑗𝑗𝑤𝑤𝑤𝑤 � maximize 𝑓𝑓𝑓𝑓�𝑋𝑋𝑋𝑋� = � 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 × 𝑝𝑝𝑝𝑝𝑖𝑖𝑖𝑖 , ratio that depends on the operation mode, p is an integer 7 ∑ 𝐹𝐹𝐹𝐹user, 𝑤𝑤𝑤𝑤𝑗𝑗𝑗𝑗 , and 𝑤𝑤𝑤𝑤 isisthe weight ofeach scheduling, 𝐹𝐹𝐹𝐹 is 𝑅𝑅𝑅𝑅the isratio that dependsofonth where 𝑅𝑅𝑅𝑅𝑝𝑝𝑝𝑝𝑖𝑖𝑖𝑖 = The the second columns columns in the arrayincontain design variables defined the 𝑗𝑗𝑗𝑗𝑖𝑖𝑖𝑖=1 The first firstandand the second the array mode, isby an𝑗𝑗𝑗𝑗=1 integer 0 or 1 for satellite, summation mode, andwhich is 0and or 1S for each satellite, andand 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 is𝑖𝑖𝑖𝑖 thethe summation of the and is 0𝑝𝑝𝑝𝑝oris1an forinteger each satellite, i is the summation of the 7 contain design variables defined by the user, which ∑ = 𝐹𝐹𝐹𝐹 𝑤𝑤𝑤𝑤 , 𝑤𝑤𝑤𝑤 is the weight of scheduling, 𝐹𝐹𝐹𝐹 is the ratio that depends where 𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖weights 𝑗𝑗𝑗𝑗 𝑗𝑗𝑗𝑗variables, product of and operation ratios corresponding to the 𝑅𝑅𝑅𝑅index represent communication. The third column represents parameters that depend design mode, 𝑝𝑝𝑝𝑝on isthe an𝑗𝑗𝑗𝑗=1 integer and iscorresponding 0 corresponding or 1 for each and is the summation ofonthft 𝑖𝑖𝑖𝑖 index weights and operation ratios tothethesatellite satellite . Theobjective objective weights and operation ratios to satellite, i . iThe represent communication. The third column represents satellite indexi. The objective function is expressed in a similar mode, 𝑝𝑝𝑝𝑝 and isinanaoperation integer form and is 0the or 0-1 1 forknapsack each satellite, andRST 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 is the i summation of tht and it plays that the role of activating communication while considering constraints. parameters depend on the design variables, and weights ratios to problem. the satellite index . The objective expressed similar and to to the expressed a similar formproblem. as as thecorresponding 0-1 knapsack problem. RST 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 𝑅𝑅𝑅𝑅correspond 𝑖𝑖𝑖𝑖 correspond form as thein0-1 knapsack RST and Si correspond to and it plays the role of activating communication while items and priority inform the knapsack However, The satellite and antenna indices represent a sequential numberingthe for satellites and antennae. The weights and ratios corresponding tovisibility the satellite index The objective to expressed aoperation similar as the 0-1 algorithm. knapsack problem. RST and 𝑅𝑅𝑅𝑅idoes 𝑖𝑖𝑖𝑖. correspond priority the knapsack algorithm. However, the TVT exactly priority in in thein knapsack algorithm. However, thethe visibility in in the TVT does notnot exactly cortc considering constraints. the visibility in the TVT does not exactly correspond to the The satellite and antenna represent sequential subscript 𝑖𝑖𝑖𝑖 indicates the rowindices of the array. In theafirst column, the satellite index is randomly defined correspond to tc in awhich similar form asthat thethat 0-1 knapsack problem. RST and 𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖does priority in the knapsack algorithm. However, the algorithm visibility in the TVT not exactly the knapsack, isreason the reason knapsack algorithm is not directly applied. knapsack, which is the the knapsack isnot theexpressed knapsack, which is the reason that thethe knapsack algorithm is directly applied. numbering for satellites and antennae. The subscript notthe directly up to the number without overlap. In the second the antenna is in applied. the index knapsack algorithm. However, the visibility in the TVT does not exactly c knapsack, which is defined the reason that the knapsack algorithm is not directly applied. i indicates the rowofofsatellites the array. In the first column, the column,priority satellite index to the number of satellitethe randomly. In is therandomly same row,defined the RSTupcorresponding to the isConstraints assigned which to the isvisibility knapsack, the reason that the knapsack algorithm is not directly applied. 4.3 Constraints 4.34.3 Constraints satellites without overlap. In the second column, the corresponding satelliterandomly. and antenna which row, represents the4.3 communication between to theto Constraints antenna index to is the defined In index, the same The RST that corresponds index variables is allocated RST that corresponds tothe thesatellite satellite index of design the variables The RST that corresponds the satellite index of of thethe design is allocated to to th

the RST corresponding to the satellite is assigned to the design variables is allocated to the visibility corresponding 4.3 Constraints The RST that toantenna the satellite index of RST themust design variables isvisibility allocated to corresponding tocorresponds the satellite antenna index. The must within visibility tim corresponding to the satellite andand index. The RST be be within thethe time visibility corresponding to the satellite and antenna to the satellite and antenna index. The RST must be within The ≤ RST that to satellite indexThe of RST the design variables is allocated corresponding the≤𝑖𝑖𝑖𝑖satellite andthe antenna index. must be within the visibility to tim index, which represents the communication between the the visibility time: 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 ≤tocorresponds 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 ≤ 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑅𝑅𝑅𝑅 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖 ≤𝑖𝑖𝑖𝑖 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖 ≤𝑖𝑖𝑖𝑖 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 . 𝑖𝑖𝑖𝑖 . satellite and antenna. The RSTs of satellites are allocated corresponding to𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 the and antenna index. The RST allocated must be within theantenna’s visibility tim 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑅𝑅𝑅𝑅 ≤ 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖 ≤ ≤is𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 (4) 𝑖𝑖𝑖𝑖reconfiguration 𝑖𝑖𝑖𝑖satellite 𝑖𝑖𝑖𝑖 . to The time is added to RSTs satellites allocated each antenna’s The reconfiguration time added thethe RSTs of of satellites to to each TV to the corresponding antennae sequentially from the onethe ground antenna and three satellites is shown in Fig. 3. The RST of a satellite is allocated within first row to the last of array. As allocation continues, 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑅𝑅𝑅𝑅 ≤ the 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 ≤must 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 . be The reconfiguration is to to thethan RSTs of allocated toRST each 𝑖𝑖𝑖𝑖 ≤ 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖time 𝑖𝑖𝑖𝑖 added The time isadded the RSTs ofsatellites satellites SET of allocated RST must smaller than SST subsequent RST to be allocT SET of𝑖𝑖𝑖𝑖reconfiguration the allocated RST be smaller thethe SST of of thethe subsequent toantenna’s be allocat visibilities, RSTs, andthereconfigurations conflict each TVT. Reconfiguration time is with required in order toallocated support other satellites after allocation to the to each antenna’s TVT, and the SET of the allocated The time to𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 the RSTs of satellites allocated to each SET of𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 allocated RST must be smaller SST of the subsequent RSTantenna’s to be allocT 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 +the 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑡𝑡 𝑅𝑅𝑅𝑅added 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑡𝑡≤ 𝑅𝑅𝑅𝑅 ≤ . the 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 +𝑘𝑘𝑘𝑘reconfiguration 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅is .than other. If a conflict occurs, the communication is not 𝑘𝑘𝑘𝑘+1 𝑘𝑘𝑘𝑘must 𝑘𝑘𝑘𝑘+1 RST be smaller the SST𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 oftothe subsequent RST to previous satellite. Communication among multiple satellites and a ground antennathan can be converted executed and the parameter corresponding to the index is allocated RST smaller than 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖𝑅𝑅𝑅𝑅must 𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑡𝑡be 𝑅𝑅𝑅𝑅 ≤ 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑘𝑘𝑘𝑘+1 . the SST of the subsequent RST to be alloc beSET allocated: 𝑘𝑘𝑘𝑘of+the a task-scheduling problem bycorresponding treating the RST and visibility as task and resource, respectively, to be assigned as zero. Otherwise, the parameter 𝑅𝑅𝑅𝑅𝑆𝑆𝑆𝑆𝑅𝑅𝑅𝑅𝑘𝑘𝑘𝑘 + 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑖𝑖𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑡𝑡𝑅𝑅𝑅𝑅 ≤ 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑘𝑘𝑘𝑘+1 . (5) to the executed index set within is given as one. The design allocated the TVT.

Figure 3. Example of support time allocation

Fig. 3. Example of support time allocation 4.1 Design variables and parameters

Design variables and parameters are obtained considering a chromosome in a genetic algorithm.

93 visibilities is regarded as an NP-hard problem. The problem of scheduling multiple satellites and

http://ijass.org

Accordingly, the designed formula is stochastic–deterministic. The overall variables are designed in an N × 3 array for N satellites: (89~100)15-125.indd 93

𝑋𝑋𝑋𝑋� = [Satellite index𝑖𝑖𝑖𝑖 , Antenna index𝑖𝑖𝑖𝑖 , pi ], pi ∈ {0, 1} .

(2)

2016-03-29 오후 7:27:34

Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016)

5. Genetic algorithm design

crossover used in this study are shown in Fig. 5. Mutation is selectively applied after the crossover depending on the initial mutation ratio. Two genes are randomly selected, and their order and the corresponding antenna are randomly changed. Then, the gene expression terms are also recalculated. The overall proposed algorithm is shown in Fig. 6. The concepts of the algorithm are based on the previous study [28] that focused on visibility allocation in the entire TVT for visibility conflict avoidance. In this study, enhanced optimal scheduling is conducted by allocating ground antenna support (actual communication time) to the visibility.

5.1 Chromosome design

The chromosome based on the design variables and the parameters described in Section 4.1 is composed of three terms: the satellite index, the antenna index, and the gene expression. The gene expression corresponds to the parameter. The diversity and feasibility of the chromosome are embodied through random generation of the index and the gene expression. The chromosome of an optimal TTC scheduling is shown in Fig. 4. The scheduling is performed sequentially with respect to the genetic expression marked as 1. Beginning from the far left algorithm edge of the chromosome, the 5. Genetic design 6. Simulation conditions th RST of the 7 satellite is allocated to the visibility between 5.1 nd Chromosome design th the 7 satellite and the 2 antenna. Although the allocation study, a ground support The chromosome on the and In the this parameters described in Section 4.1 is resolution algorithm is continuously performed from left based to right, it design causesvariables a for visibility conflicts is proposed based on KOMPSATs. resource conflict at the 13th satellite; thus, communication composed of three terms: the satellite index, the antenna index, and the gene expression. The gene However, the visibilities of currently operation KOMPSATs is not performed with the 13th satellite. The schema in scarcely conflict with other; are therefore, the conflicts expression corresponds to the parameter. The diversity and feasibility of the each chromosome the genetic algorithm is defined as the gene order in this are not of sufficient complexity to require an optimization algorithm. embodied through random generation of the index and the gene expression. The chromosome of an algorithm. However, because more satellites will be deployed optimal TTC scheduling is shown in Fig. 4. The scheduling performed with respect to in theis future, wesequentially assume numerous sun-synchronous orbit 5.2 Crossover and mutation satellites whose specifications areof randomly generated the genetic expression marked as 1. Beginning from the far left edge of the chromosome, the RST If a general crossover method is used, the same satellites analogous to the current KOMPSATs. Three terrestrial the 7th satellite is allocated to the visibility between the 7th satellite and the 2nd antenna. Although the will overlap. A satellite should exist without overlap to antennae are assumed according to Korean ground control allocation is continuously performed from left to right, it causes a resource conflictbetween at the 13th satellite; maintain diversity and feasibility within a chromosome. stations. Visibilities the satellites and ground Therefore, a cycle crossover is applied and the with gene antennae are calculated through the thus, communication is not performed the 13th satellite. The schema in the genetic algorithm is STK (Fig. 7). Fig. 8 expression term is recalculated sequentially. Examples of the shows the visibility between the 1st antenna and satellites defined as the gene order in this algorithm. Satellite 7 index Antenna 2 index Genetic 1 expression

14

6

3

8

12 15 11 13

9

10

5

18

2

19 17

4

1

16 20

2

3

1

1

2

3

3

3

3

1

3

2

2

2

2

2

2

2

1

1

1

1

1

1

1

1

0

0

1

1

1

0

0

0

0

0

0

0

Figure 4. Chromosome in TTC optimal scheduling

Fig. 4. Chromosome in TTC optimal scheduling 5.2 Crossover andstmutation

2nd iteration

1 iteration

If1 a1 general17 crossover method is 17 used, the same satellites A satellite should 7 1 1 will overlap. 17 1 1 7 1 1 17 exist 1 1 1 1 7 1 1 1 1

7 10 11 4

1 1

11 3 1

10 1 1 11 1 1

11 3 1 19 3 1

11 3 1 19 3 1

10 1 1 11 1 1

13 2 1 17 1 0

9 3 0 13 2 1

13 2 1 17 1 0

9 3 0 13 2 1

16 2 1 3 1 0

3 1 0 16 3 0

20 1 0 5 3 0 10 1 0

8 1 0 7 3 0 14 1 0

1 1 19 3 1 without overlap to maintain diversity and feasibility within a chromosome. Therefore, a4 cycle 4 3 1 14 1 1 14 1 1 3 1 3 1 14 1 1 13 2 1

9

3 0

16 2 1

3

1 0

20 1 0

8

1 0

5 14 18 19

3 1 2 2

0 0 0 1

7 10 5 12

3 1 3 3

0 0 0 1

15 9 8 12

2 3 3 2

0 1 1 0

20 6 18 4

1 2 2 1

0 0 1 0

6 1

1 0 2 0

13 2 1

9 3 0

17 1 0 13 2 1 1 0 13 2 1 crossover is applied and17 the gene expression term is recalculated Examples of the 2 2 0 2sequentially. 1 1 2 2 0 2 1 1 3 1 0 0 crossover used16 in3 this study are shown in Fig. 5.

Parent 1

Fig. 5. Cycle crossover application

1 2 0 15 1 0 Parent 2

20 1 0 5 3 0

8 1 0 7 3 0

18 2 0

5 3 0

15 2 0 9 3 1 8 3 1

20 1 0 6 2 0 18 2 1

6 1

1 2 0 15 1 0

1 0 2 0

Offspring 1

Offspring 2

20 1 0 5 3 0

8 7

14 1 0 18 2 0

10 1 0 5 3 0

19 15 9 8

12 20 6 18

2 2 3 3

1 0 1 1

1 0 3 0

3 1 2 2

1 0 0 1

12 2 0 6 1 0 1 2 0

4 1 0 1 2 0 15 1 0

Parent 1

Parent 2

18 2 0

5

15 2 0 9 3 1

20 1 0 6 2 0

8

3 1

18 2 1

6 1

1 0 2 0

1 2 0 15 1 0

Offspring 1

3 0

Offspring 2

Figure 5. Cycle crossover application

Mutation is selectively applied after the crossover depending on the initial mutation ratio. Two genes are randomly selected, and their order and the corresponding antenna are randomly changed. DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89 94 Then, the gene expression terms are also recalculated. The overall proposed algorithm is shown in Fig. 6. The concepts of the algorithm are based on the previous study [28] that focused on visibility allocation in the entire TVT for visibility conflict avoidance. In this study, enhanced optimal (89~100)15-125.indd 94

scheduling is conducted by allocating ground antenna support (actual communication time) to the

2016-03-29 오후 7:27:35

Junghyun Lee Genetic algorithm–based scheduling for ground support of multiple satellites and antennae considering .....

synchronous orbit satellites are assumed, and each has local time and altitude information. Their indices are numbered according to the AOS of the satellites. Based on the considerations described in Section 3, numerical values are given in sequence. Individual satellites require the support of ground antennae for the duration of their RST during the visibility.

and the resulting conflicts during a 30 min simulation period from 04:00:00.000 to 04:30:00.000 UTCG on Feb 21, 2014. The satellites contacting the three ground antennae and generating visibility conflicts during the period are selected (Table 4).

6.1 Satellites Twenty satellites are assumed for this simulation. The orbit information of the satellites and their scheduling considerations are summarized in Table 4. Sun-

6.2 Antennae Three ground antennae are assumed. Their details are

Initial population

Genetic operation

6. Simulation conditions In this study, a ground support resolution algorithm for visibility conflicts is proposed based on KOMPSATs. However, the visibilities of currently operation KOMPSATs scarcely conflict with each other; therefore, the conflicts are not of sufficient complexity to require an optimization algorithm. However, because more satellites will be deployed in the future, we assume numerous sunsynchronous orbit satellites whose specifications are randomly generated analogous to the current KOMPSATs. Three terrestrial antennae are assumed according to Korean ground control stations. Visibilities between the satellites and ground antennae are calculated through the STK (Fig. 7). Fig. 8 shows the visibility between the 1st antenna and satellites and the resulting conflicts during a 30 min simulation period from 04:00:00.000 to 04:30:00.000 UTCG on Feb 21, 2014. The satellites contacting the three ground antennae and generating visibility conflicts during the period are selected Figure 6. Proposed genetic algorithm procedure Fig. 6. Proposed genetic algorithm procedure (Table 4).

Fig. 7. STK modeling

Figure 7. STK modeling

95

(89~100)15-125.indd 95

http://ijass.org

2016-03-29 오후 7:27:35

Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016)

given in Table 5, where the antenna index indicates the station number and latitude and longitude represent the position. As this study focuses on the Korean satellite environment,

the ground antennae are assumed to be located in Seoul, Daejeon, and Goheung. The cone half-angles of all antennae are defined as 60°.

Fig. 8. Visibility between 1st antenna and satellites Figure 8. Visibility between 1st antenna and satellites Table 4. Scheduling considerations of satellites

Table 4. Scheduling considerations of satellites Profit

Contact interval (hours)

Satellite performance

5

30

2

2

40

1

Satellite index

Priority

1 2

6.1 Satellites

Emerg ency

Remaining resources (%)

RST (s)

Altitude

High

8

114.3

698.7

11:41

Low

28

149.7

684.6

12:53

Local time (descending node)

Twenty satellites are assumed for this simulation. The orbit information of the satellites and their 3 7 10 5 Low 52 157.6 673.2 11:43 scheduling considerations are summarized in Table 4. Sun-synchronous orbit satellites are assumed, 4

9

30

3

High

65

158.0

675.6

12:03

5 has local 6 8 Middle 128.6 670.6 and each time and 40 altitude information. Their indices25are numbered according12:13 to the AOS of 6

4

100

20

10

Low

22

181.8

680.2

12:48

the satellites. described in Section 3,137.5 numerical are given in 7 3Based on the20considerations 5 High 85 671.0 values 12:17 8

1

50

7

Low

82

151.1

699.9

12:29

9

3

20

2

Low

15

175.7

687.6

12:37

20

4

Low

45

169.8

684.2

12:38

30

6

Low

23

165.0

698.5

12:22

10

6

High

82

108.5

685.1

12:44

sequence. Individual satellites require the support of ground antennae for the duration of their RST 7 during10the visibility. 11

8

12

2

100

13

7

30

1

Low

84

159.5

686.8

12:41

14

8

40

7

Middle

23

101.5

673.0

12:30

15

4

30

2

Low

15

111.6

687.7

12:24

16

10

50

1

Middle

86

127.1

693.5

13:13

17

2

10

6

Low

55

109.2

697.5

12:58

18

2

30

7

Middle

12

179.2

693.3

12:19

19

1

40

5

Middle

61

171.9

674.1

12:31

20

6

50

1

Low

65

112.4

696.6

12:44

6.2 Antennae DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89 96 Three ground antennae are assumed. Their details are given in Table 5, where the antenna index indicates the station number and latitude and longitude represent the position. As this study focuses on the Korean satellite environment, the ground antennae are assumed to be located in Seoul, Daejeon, (89~100)15-125.indd 96

and Goheung. The cone half-angles of all antennae are defined as 60°.

2016-03-29 오후 7:27:36

11

8

12

2

100

30

6

Low

23

165.0

698.5

12:22

10

6

High

82

108.5

685.1

12:44

13

7

30

1

Low

84

159.5

686.8

12:41

14

8

40

7

Middle

23

101.5

673.0

12:30

15

4

30

2

Low

15

111.6

687.7

12:24

16

10

50

1

Middle

86

127.1

693.5

13:13

17

2

10

6

Low

55

109.2

697.5

12:58

18

2

30

7

Middle

12

179.2

693.3

12:19

5

Middle

61

171.9

674.1

1

Low

65

112.4

696.6

19

1

20

6

Junghyun Lee Genetic algorithm–based scheduling for ground support of multiple satellites and antennae considering ..... 40

6.3

50 Visibility

12:31

6.412:44 Reconfiguration

time

The visibilities between the twenty satellites and the The reconfiguration time is assumed as 50 seconds. three ground antennae are listed in Table 6 and calculated In terms of task scheduling, the reconfiguration time is 6.2 Antennae with respect to corresponding satellites and antennae as allocated between the RSTs. Three ground antennae are assumed. Theirthe details given in Table 5, where the difference between LOSareand AOS, which causesthe antenna index a conflict between satellites. The TVT is expressed as the 7. Results indicates the station number and latitude and longitude represent the position. As this study focuses on difference between the last LOS and the first AOS. The the Korean satellite environment, the ground are antennae assumed to can be located in Seoul, Daejeon, TVT is the period duringantennae which the provide 7.1 Application to operation modes support.

and Goheung. The cone half-angles of all antennae are defined as 60°.

The proposed algorithm is applied to two operation modes: TTC and PL. In the case of TTC, the emphasis is on telemetry Table 5. Ground antenna Table 5.positions Ground antenna positions and tele-command communication for maintaining the SOH 6.3 Visibility Antenna of the satellites. The commands that have a large influence Latitude Longitude Cone half-angle index The visibilities between the twenty satellites and the three ground antennae are listed in Table 6 and on SOH, such as urgency and contact interval, are given 1 37.6° N 127.0° E 60° calculated with respect to corresponding satellites andlarger antennae as the difference between LOScase of PL, the focus is weights than other data. the In the 2 36.4° N 127.4° E 60° onTVT obtaining the asinherent payload such as image data and and AOS, which causes a conflict between satellites. The is expressed the difference between 3 34.6° N 127.3° E 60° its benefits; therefore, satellite performance and remaining the last LOS and the first AOS. The TVT is the period during which the antennae can provide support.

6. Visibility between ground antennae and satellites Table 6. Visibility between ground antennae andTable satellites Antenna index

1

2

3

Satellite index

AOS

LOS

LOS

LOS

(mm:ss)

(mm:ss)

Visibility (s)

AOS

(mm:ss)

Visibility (s)

AOS

(mm:ss)

(mm:ss)

(mm:ss)

1

20:04.4

22:49.1

164.7

20:40.8

22:46.5

125.7

21:28.3

22:53.2

85.0

2

16:13.2

19:23.8

190.5

16:24.0

19:50.9

206.8

16:49.8

20:21.9

212.1

3

11:49.2

14:49.4

180.2

12:21.2

14:51.3

150.1

13:01.8

15:04.8

123.0

4

12:12.6

16:34.2

261.6

12:34.3

16:47.6

253.3

13:04.8

17:11.9

247.1

5

10:37.0

14:59.8

262.8

10:58.3

15:13.6

255.3

11:28.4

15:38.2

249.8

6

14:36.3

18:07.3

211.0

14:48.5

18:32.7

224.3

15:14.5

19:03.3

228.8

7

10:49.2

15:24.6

275.4

11:07.5

15:42.3

274.8

11:35.6

16:09.5

273.8

8

20:03.6

24:49.3

285.7

20:21.3

25:07.8

286.4

20:49.3

25:35.5

286.1

9

16:24.9

20:49.2

264.3

16:40.4

21:10.6

270.1

17:07.4

21:39.7

272.3

10

15:23.8

19:41.6

257.7

15:38.9

20:03.4

264.5

16:05.8

20:32.8

267.0

11

19:32.0

24:17.5

285.5

19:50.9

24:34.4

283.4

20:19.6

25:01.1

281.5

12

15:53.8

19:52.9

239.2

16:07.6

20:16.5

248.8

16:34.1

20:46.4

252.3

13

16:18.0

20:30.3

252.4

16:32.7

20:52.8

260.2

16:59.4

21:22.4

263.0

14

11:42.0

16:05.8

263.8

11:57.9

16:26.5

268.6

12:25.0

16:55.3

270.3

7.1 15 Application to operation 16:21.7 20:52.4 modes 270.7

16:37.8

21:13.0

275.2

17:05.0

21:41.8

276.8

time is allocated between the RSTs.

7. Results

Visibility (s)

16

19:12.1

22:10.7

178.6

19:22.0

22:39.0

197.0

19:47.6

23:10.4

202.8

17

20:23.0

23:33.2

190.3

20:33.6

24:00.7

207.1

20:59.4

24:31.9

212.6

The proposed algorithm is applied to two operation modes: TTC and PL. In the case of TTC, the 18 22:40.3 266.1 18:29.7 23:01.7 for272.0 18:56.7 274.1 emphasis is18:14.2 on telemetry and tele-command communication maintaining the SOH23:30.8 of the satellites. 19

12:04.1

16:27.0

262.9

12:19.9

16:47.8

268.0

12:46.9

17:16.7

269.8

20

19:25.1

23:41.6

256.6

19:39.8

24:04.1

264.3

20:06.6

24:33.7

267.1

The commands that have a large influence on SOH, such as urgency and contact interval, are given larger weights than First AOS (hh:mm:ss)

other data. In10:37.0 the case of PL, the focus is on obtaining the inherent payload11:28.4 such 10:58.3

Last LOS (hh:mm:ss)

24:49.3

25:07.8

25:35.5

852.3

849.5

847.1

as image data and its benefits; therefore, satellite performance and remaining resources received larger

TVT (s)

weights. The relative importance of the scheduling considerations is summarized in Table 7.

Table 7. Ratio depending6.4 onReconfiguration the operation mode time

Table 7. Ratio depending on the operation mode

User Contact Support The reconfiguration time is Emergency assumed as Profit 50 seconds. In terms of task Satellite scheduling,Remaining the reconfiguration Case Total interval

time

performance

resources

TTC

priority 1

4

0.25

4

0.25

0.25

0.25

10

PL

1

0

3

0.25

0.5

3

2.25

10

7.2 Simulation results A simulation is executed for the modeling and 97 the genetic algorithm presented in Section 5 and 6,

http://ijass.org

respectively. The simulation conditions of the genetic algorithm are given in Table 8. The number of populations and generations are 100 and 500, respectively. The selection pressure of the roulette (89~100)15-125.indd 97

selection is set at 5. The two operation modes (TTC and PL) are analyzed separately, the results of

2016-03-29 오후 7:27:36

Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016)

resources received larger weights. The relative importance of the scheduling considerations is summarized in Table 7.

algorithm presented in Section 5 and 6, respectively. The simulation conditions of the genetic algorithm are given in Table 8. The number of populations and generations are 100 7.2 Simulation results and described 500, respectively. The selection pressure of the roulette reasonable for real applications (Table 9). The chromosome in Fig. 4 is implemented as the reasonable for real applications (Table 9). The chromosome described in Fig. 4 is implemented as the selection is set at 5. The two operation modes (TTC and PL) A simulation is executed for the modeling and the genetic scheduling type in Table 9. scheduling type in Table 9.

Table 8. Simulation conditions Table 8. Simulation conditions Hardware and simulation tools Genetic condition Hardware and simulation tools Genetic condition Processor Intel® Core™ i7-4790 CPU @ 3.60 GHz Number of populations 100 Processor Intel® Core™ i7-4790 CPU @ 3.60 GHz Number of populations Memory (RAM) 16 GB Number of generations 500 Memory (RAM) 16 GB Number of generations Mutation ratio 5% Simulation tools MATLAB 2011 a Mutation ratio Selection pressure 5 Simulation tools MATLAB 2011 a Selection pressure

Table 8. Simulation conditions

Table 9. Scheduling results and running time

100 500 5% 5

Table 9. Scheduling results and running time Table 9. Scheduling results and running time

(a) TTC mode (a) TTC mode

(b) PL mode (b) PL mode

Antenna - 1 Antenna - 1 Antenna - 1 Antenna - 1 Satellite Satellite AOS EOS SST SET Objective AOS EOS SST SET index Satellite index Satellite AOS EOS SST SET Objective AOS EOS SST index index 3 11:49.2 14:49.4 11:49.2 14:26.8 80.0 4 12:12.6 16:34.2 12:12.6 14:50.6 3 11:49.2 14:49.4 11:49.2 14:26.8 80.0 4 12:12.6 16:34.2 12:12.6 10 15:23.8 19:41.6 15:23.8 18:13.6 62.4 2 16:13.2 19:23.8 16:13.2 18:43.0 10 15:23.8 19:41.6 15:23.8 18:13.6 62.4 2 16:13.2 19:23.8 16:13.2 8 20:03.6 24:49.3 20:03.6 22:34.7 101.1 1 20:04.4 22:49.1 20:04.4 21:58.6 8 20:03.6 24:49.3 20:03.6 22:34.7 101.1 1 20:04.4 22:49.1 20:04.4 Sum 243.6 Sum Sum 243.6 Antenna - 2 Antenna - 2 Antenna - 2 Antenna - 2 Satellite Total Satellite AOS EOS SST SET AOS EOS SST SET index Satellite objective Total index Satellite AOS EOS SST SET AOS EOS SST index objective index 7 11:07.5 15:42.3 11:07.5 13:25.0 83.4 19 12:19.9 16:47.8 12:19.9 15:11.8 7 11:07.5 15:42.3 11:07.5 13:25.0 83.4 19 12:19.9 16:47.8 12:19.9 14 11:57.9 16:26.5 14:15.0 15:56.5 115.1 12 16:07.6 20:16.5 16:07.6 17:56.1 14 11:57.9 16:26.5 14:15.0 15:56.5 115.1 12 16:07.6 20:16.5 16:07.6 12 16:07.6 20:16.5 16:46.5 18:35.0 80.6 15 16:37.8 21:13.0 18:46.2 20:37.8 12 16:07.6 20:16.5 16:46.5 18:35.0 80.6 15 16:37.8 21:13.0 18:46.2 18 18:29.7 23:01.7 19:25.0 22:24.2 106.5 20 19:39.8 24:04.1 21:27.8 23:20.2 18 18:29.7 23:01.7 19:25.0 22:24.2 106.5 20 19:39.8 24:04.1 21:27.8 Sum 385.6 Sum Sum 385.6 Antenna - 3 Antenna - 3 Antenna - 3 Antenna - 3 Satellite Total Satellite AOS EOS SST SET AOS EOS SST SET index Satellite objective Total index Satellite AOS EOS SST SET AOS EOS SST index objective index 5 11:28.4 15:38.2 11:28.4 13:37.0 112.3 5 11:28.4 15:38.2 11:28.4 13:37.0 5 11:28.4 15:38.2 11:28.4 13:37.0 112.3 5 11:28.4 15:38.2 11:28.4 6 15:14.5 19:03.3 15:14.5 18:16.3 500.7 14 12:25.0 16:55.3 14:27.0 16:08.5 6 15:14.5 19:03.3 15:14.5 18:16.3 500.7 14 12:25.0 16:55.3 14:27.0 15 17:05.0 21:41.8 19:06.3 20:58.0 46.0 9 17:07.4 21:39.7 17:07.4 20:03.1 15 17:05.0 21:41.8 19:06.3 20:58.0 46.0 9 17:07.4 21:39.7 17:07.4 11 20:19.6 25:01.1 21:48.0 24:33.0 487.4 16 19:47.6 23:10.4 20:53.2 23:00.2 11 20:19.6 25:01.1 21:48.0 24:33.0 487.4 16 19:47.6 23:10.4 20:53.2 Sum 1146.5 Sum Sum 1146.5 Number of allocated satellites 11 Number of allocated satellites Number of allocated satellites 11 Number of allocated satellites Total objective 1775.6 Total objective Total objective 1775.6 Total objective Runtime (seconds) 55.1 Runtime (seconds) Runtime (seconds) 55.1 Runtime (seconds)

(a)

TTC mode

Fig. 9. Fitness variations according to generation

(a)

Objective SET Objective 268.1 14:50.6 268.1 227.1 18:43.0 227.1 305.1 21:58.6 305.1 800.3 Sum 800.3 Total objective SET Total objective 246.0 15:11.8 246.0 190.9 17:56.1 190.9 214.2 20:37.8 214.2 235.7 23:20.2 235.7 886.8 Sum 886.8 Total objective SET Total objective 280.6 13:37.0 280.6 284.1 16:08.5 284.1 180.9 20:03.1 180.9 270.4 23:00.2 270.4 1016.1 Sum 1016.1 11 11 2703.2 2703.2 57.5 57.5

PL mode

Figure 9. Fitness variations according to generation

DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89 8. Conclusion

98

We proposed a near-global optimal scheduling algorithm for ground support allocation of multiple satellites and ground antennae via a genetic algorithm, focusing on KOMPSATs. The proposed algorithm not only provides maximum benefit to the user but also enhances applicability to actual (89~100)15-125.indd 98

operations by considering various scheduling factors and operation modes. Because satellite–antenna

2016-03-29 오후 7:27:37

Junghyun Lee Genetic algorithm–based scheduling for ground support of multiple satellites and antennae considering .....

spacecraft missions: a GA model and its results”, Aerospace and Electronic Systems, IEEE Transactions on, Vol. 45, 2009, pp. 950-964. [2] Buckreuss, S., Mühlbauer, P., Mittermayer, J., Balzer, W. and Werninghaus, R., “The TerraSAR-X ground segment”, EUSAR 2006, 2006. [3] Colin, H., Gian Paolo, C., Vicente, N. and Nuno, S., “ESOC Ground Segments - The Future?”, in SpaceOps 2008 Conference, ed: American Institute of Aeronautics and Astronautics, 2008. [4] Noguero, J., Garcia Julian, G. and Beech, T. W., “Mission control system for Earth observation missions based on SCOS-2000”, in Aerospace Conference, 2005 IEEE, 2005, pp. 4088-4099. [5] Alessandro, L., Daniele Di, N., Sylvie, H., Mauro, P., Bargellini, P. and Tiago, N., “Mission Automation Infrastructure Tools at ESOC”, in SpaceOps 2008 Conference, ed: American Institute of Aeronautics and Astronautics, 2008. [6] Lee, B., Lee, J., Park, J., Kim, H. and Kim, J., “Implementation of the Mission Scheduling and Command Planning Functions for the KOMPSAT -2 Mission Control Element”, in The Korean Society For Aeronautical And Space Sciences, 2003, pp. 707-710. [7] Jung, W., Lee, B., Lee, S. and Kim, J., “Mission Control System for KOMPSAT-2 Operations”, IEIC Technical Report (Institute of Electronics, Information and Communication Engineers), Vol. 106, 2006, pp. 169-176. [8] Kim, H. and Lee, J., “Simulation for the Mission Planning of the KOMPSAT MCE”, in The Korean Society For Aeronautical And Space Sciences, 1998, pp. 564-567. [9] Kim, H.-O., Kim, H.-S., Lim, H.-S. and Choi, H.-J., “Space-Based Earth Observation Activities in South Korea [Space Agencies]”, Geoscience and Remote Sensing Magazine, IEEE, Vol. 3, 2015, pp. 34-39. [10] Jain, A. S. and Meeran, S., “Deterministic job-shop scheduling: Past, present and future”, European journal of operational research, Vol. 113, 1999, pp. 390-434. [11] Sule, D. R., Industrial Scheduling: PWS Publishing Company, 1997. [12] Cormen, T. H., Leiserson, C. E., Rivest, R. L. and Stein, C., Introduction to algorithms: The MIT press, 2001. [13] Martello, S. and Toth, P., Knapsack problems: algorithms and computer implementations: John Wiley & Sons, Inc., 1990. [14] Baek, S., Cho, K., Lee, D. and Kim, H., “A Comparison of Scheduling Optimization Algorithm for the Efficient Satellite Mission Scheduling Operation”, in The Korean Society For Aeronautical And Space Sciences, 2010, pp. 48-57. [15] Baek, S., Han, S., Cho, K., Lee, D., Yang, J., Bainum, P. M., et al., “Development of a scheduling algorithm and GUI

are analyzed separately, the results of which are summarized in Table 9. At-risk satellites, i.e., the 6th and the 11th are allocated in TTC mode, not PL mode. The 5th, 6th, 8th, 14th, and 18th satellites, which have not made contact for more than seven hours, are all allocated in TTC mode. The 1st, 9th, and 15th satellites, which have the fewest remaining resources, are allocated in PL mode. These patterns represent the effect of the operation mode (Table 7). Fig. 9 depicts the fitness, which characterizes the genetic algorithm, as the generations progress. The fitness result shows that the algorithm performed suitably in both operation modes. The simulation was conducted in MATLAB 2011a, and the hardware conditions are listed in Table 8. The computational runtime was approximately 50 s for both TTL and PL modes, which is reasonable for real applications (Table 9). The chromosome described in Fig. 4 is implemented as the scheduling type in Table 9.

8. Conclusion We proposed a near-global optimal scheduling algorithm for ground support allocation of multiple satellites and ground antennae via a genetic algorithm, focusing on KOMPSATs. The proposed algorithm not only provides maximum benefit to the user but also enhances applicability to actual operations by considering various scheduling factors and operation modes. Because satellite–antenna scheduling is an NP-hard problem, an optimization formula and a genetic algorithm were specially designed. We designed decision variables and a chromosome as an array type to satisfy diversity, feasibility, and convergence. The algorithm was verified through simulation. The convergence and near-optimality are evident in fitness plots. By applying different relative weighting depending on the operation mode, i.e., TTC and PL, we showed that the proposed method is applicable to various operation modes using examples.

Acknowledgement This research was supported by the EDISON Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (2014M3C1A6038801).

References [1] Bagchi, T. P., “Near optimal ground support in multi-

99

(89~100)15-125.indd 99

http://ijass.org

2016-03-29 오후 7:27:37

Int’l J. of Aeronautical & Space Sci. 17(1), 89–100 (2016)

Multi-Satellite Tracking”, in ASME Conference Proceedings, 2011, pp. 727-734. [28] Lee, J., Wang, S., Chung, D., Hyun, C., Choi, S., Ko, K., Ahn, H. and Jung, O., “Visibility conflict resolution for multiple antennae and multi-satellites via genetic algorithm”, in Aerospace Conference, 2013 IEEE, 2013, pp. 1-10. [29] Corrao, G., Falone, R., Gambi, E. and Spinsante, S., “Ground station activity planning through a multi-algorithm optimisation approach”, in Satellite Telecommunications (ESTEL), 2012 IEEE First AESS European Conference on, 2012, pp. 1-6. [30] Xhafa, F., Sun, J., Barolli, A., Takizawa, M. and Uchida, K., “Evaluation of genetic algorithms for single ground station scheduling problem”, in Advanced Information Networking and Applications (AINA), 2012 IEEE 26th International Conference on, 2012, pp. 299-306. [31] Spangelo, S., Cutler, J., Gilson, K. and Cohn, A., “Optimization-based scheduling for the single-satellite, multi-ground station communication problem”, Computers & Operations Research, Vol. 57, 2015, pp. 1-16. [32] Sun, J. and Xhafa, F., “A genetic algorithm for ground station scheduling”, in Complex, Intelligent and Software Intensive Systems (CISIS), 2011 International Conference on, 2011, pp. 138-145. [33] Xhafa, F., Sun, J., Barolli, A., Biberaj, A. and Barolli, L., “Genetic algorithms for satellite scheduling problems”, Mobile Information Systems, Vol. 8, 2012, pp. 351-377. [34] Lee, S., Jung, W. C. and Kim, J., “Task scheduling algorithm for the communication, ocean, and meteorological satellite”, ETRI journal, Vol. 30, 2008, pp. 1-12. [35] Patterson, J. H., “A comparison of exact approaches for solving the multiple constrained resource, project scheduling problem”, Management science, Vol. 30,1984, pp. 854-867. [36] Shen, W., “Distributed manufacturing scheduling using intelligent agents”, Intelligent Systems, IEEE, Vol. 17, 2002, pp. 88-94. [37] Smith, S. F., “Is scheduling a solved problem?”, in Multidisciplinary Scheduling: Theory and Applications, ed: Springer, 2005, pp. 3-17. [38] Agnese, J., Bataille, N., Bensana, E., Blumstein, D. and Verfaillie, G., “Exact and approximate methods for the daily management of an earth observation satellite”, in Proc. of the 5th ESA Workshop on Artificial Intelligence and Knowledge Based Systems for Space, 1995.

for autonomous satellite missions”, Acta Astronautica, Vol. 68, 2011, pp. 1396-1402. [16] Han, S., Baek, S., Jo, S., Cho, K., Lee, D. and Kim, H., “Optimization of the Satellite Mission Scheduling Using Genetic Algorithms”, in The Korean Society For Aeronautical And Space Sciences, 2008, pp. 1163-1170. [17] Dishan, Q., Chuan, H., Jin, L. and Manhao, M., “A Dynamic Scheduling Method of Earth-Observing Satellites by Employing Rolling Horizon Strategy”, The Scientific World Journal, Vol. 2013, 2013. [18] Frank, J., Jonsson, A., Morris, R. and Smith, D., “Planning and scheduling for fleets of earth observing satellites”, in Proceedings of the sixth international symposium on artificial intelligence, robotics, automation and space, 2001. [19] Lin, W.-C. and Liao, D.-Y., “A tabu search algorithm for satellite imaging scheduling”, in Systems, Man and Cybernetics, 2004 IEEE International Conference on, 2004, pp. 1601-1606. [20] Lin, W.-C., Liao, D.-Y., Liu, C.-Y. and Lee, Y.-Y., “Daily imaging scheduling of an earth observation satellite”, Systems, Man and Cybernetics, Part A: Systems and Humans, IEEE Transactions on, Vol. 35, 2005, pp. 213-223. [21] Globus, A., Crawford, J., Lohn, J. and Pryor, A., “Scheduling earth observing satellites with evolutionary algorithms”, in Conference on Space Mission Challenges for Information Technology (SMC-IT), 2003. [22] Globus, A., Crawford, J., Lohn, J. and Pryor, A., “A comparison of techniques for scheduling earth observing satellites”, in AAAI, 2004, pp. 836-843. [23] Arkali, G., Dawande, M. and Sriskandarajah, C., “Scheduling support times for satellites with overlapping visibilities”, Production and Operations Management, Vol. 17, 2008, pp. 224-234. [24] Soma, P., Venkateswarlu, S., Santhalakshmi, S., Bagchi, T. and Kumar, S., “Multi-satellite scheduling using genetic algorithms”, ISTRAC/ISRO, SpaceOps, 2004. [25] Rao, J., Soma, P. and Padmashree, G., “Multi-satellite scheduling system for LEO satellite operations”, SpaceOps, Thokyo, Japan, 2b002, 1998. [26] Castaing, J., “Scheduling Downloads for MultiSatellite, Multi-Ground Station Missions”, presented at the Small Satellite Conference 2014, Logan, Utah, 2014. [27] Yun, S.-H., Ahn, H.-S., Park, S.-J., Jung, O.-C., and Chung, D.-W., “Ground Antenna Scheduling Algorithm for

DOI: http://dx.doi.org/10.5139/IJASS.2016.17.1.89

(89~100)15-125.indd 100

100

2016-03-29 오후 7:27:37