Real Time Scheduling of Data Transmission Sessions in a

Microsatellites Swarm and Ground Stations Network Based on

Multi-Agent Technology

P. Skobelev

1,3

, E. Simonova

1,3

, A. Ivanov

, I. Mayorov

, V. Travin

and A. Zhilyaev

1,3

Samara State Aerospace University, 34, Moskovskoye Shosse, Samara, Russia

École Polytechnique Fédérale de Lausanne, Route Cantonale, 1015 Lausanne, Switzerland

SEC “Smart Solutions”, Ltd, 17, Moscovskoe Shosse, office center “Vertikal”, office 1201, Samara, Russia

Keywords: Microsatellite, Ground Station, Multi-Agent Technology, Communication Session, Ontology, Data-Stream

Scheduling, Agent, Real Time.

Abstract: The problem of designing an effective models and methods for data transmission between group of

microsatellites and network of ground stations in the dynamically changing environment is considered.

Multi-agent technology for solving the problem by adaptive resource allocation and scheduling is proposed.

It is shown that solution of the considered complex problem evolutionary emerges from interaction and

trade-offs of many agents which continuously self-organize themselves and change decisions to improve

their objectives and the objectives of the system as a whole. The advantages of multi-agent solution are high

adaptability, flexibility and efficiency of services. The main classes of agents, ontology of problem domain,

interaction protocols, results of first experiments with system prototype and key benefits of proposed system

are discussed.

1 INTRODUCTION

Recent achievements in miniaturization of satellites

components made easy access to space for SME and

universities. As the result, a new class of satellites

called small-scale have appeared. The current

classification of such satellites has formed by their

weight characteristics: pico-satellites - under 1 kg,

nano-satellites - from 1 to 10 kg, microsatellites -

from 10 to 100 kg, mini-satellites - from 100 to 500

kg and small SC - from 500 to 1000 kg.

New trend of small satellites industry is

characterized by designing microsatellites swarms,

capable of continuous earth surface monitoring,

resolve navigation and telecommunication issues,

etc. The small satellites swarms can dramatically

change the market of space systems and provide new

low-cost services comparing to the ones that are

implemented today (Zinchenko, 2011, Global

Navigation Satellite Systems, 2012). On the other

hand, increased number of microsatellites in orbit

results in overload of management systems and the

necessity to process large amounts of target

information. Under these circumstances, a problem

of scheduling the timely data transfer from

microsatellites to the ground stations becomes one of

the most urgent.

In this paper the multi-agent technology for

solving the problem by adaptive resource allocation

and scheduling is proposed. It will be shown that

solution of the considered complex problem could

evolutionary emerge from interaction and trade-offs

of many agents which continuously change

decisions to improve their objectives and the

objectives of the system as a whole.

The paper is organized as follows. First part will

formalize the problem statement and give method of

scheduling based on multi-agent technology. The

second part will consider the architecture of multi-

agent system for problem solving including specific

classes of agents and protocols of their interaction.

The third part will present results of first

experiments with software prototype of the system.

At the end the key benefits of proposed system

and future steps will be discussed.

153

Skobelev P., Simonova E., Ivanov A., Mayorov I., Travin V. and Zhilyaev A..

Real Time Scheduling of Data Transmission Sessions in a Microsatellites Swarm and Ground Stations Network Based on Multi-Agent Technology.

DOI: 10.5220/0005034301530159

In Proceedings of the International Conference on Evolutionary Computation Theory and Applications (ECTA-2014), pages 153-159

ISBN: 978-989-758-052-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

2 PROBLEM STATEMENT

Let’s consider a swarm of microsatellites that

belongs to different types of users (e.g. several

universities, SME, ERS operators, etc.) and focused

on getting the information from microsatellites and

on data transceiving to the ground stations.

Microsatellites swarm functionality can change

in time (some of them break down, the new ones are

launched, etc.). Each microsatellite can have

constraints of technical, organizational, financial or

of any other character on the data transmission to the

ground stations that belong to different developers.

Therefore there is a group of available stations for

data transfer and the group of temporary available

stations. Let’s assume that all ground stations have

an ability to transmit the received information to

another station with the use of Internet. Data

transmission task is a task to transmit the specified

data amount from one particular microsatellite in a

given time period. Data transmission to the ground

station should be proceeded by a communication

sessions which need to be scheduled properly. It is

necessary for the implementation of some

preliminary work on the station, including the

computation of target destinations for antenna

aiming to a certain satellite, preparation of

engineering software tools for processing of

incoming information, etc.

The main idea of the suggested development is to

design a method of real time scheduling for

microsatellites data transmittion sessions to ground

stations system with dynamic constraints. Duration

of communication sessions between microsatellites

and ground stations is one of the key factors that

affect monitoring performance and data delivery

efficiency. Therefore, data delivery to the ground

stations optimization influences directly satellites

efficiency.

The main objective for sarm is to provide the

efficient data transmittion from microsatellites at the

required time with the minimum delay from the

moment of on-board data receipt. At the same time,

system should adaptively correct schedule for each

station, considering the unpredictable events:

failures on a satellite, station equipment failure, new

VIP user request on data receipt, etc. If unexpected

events have occurred on one of the stations, its tasks

should be redistributed between other stations of the

network.

The considered problem of data transmission

sessions scheduling between many microsatellites

and ground stations can be more formally

formulated in the following way. There must be

provided data flow Ф maximization in the

microsatellites system N on the time horizon T for

M stations:

,)()()( dttscheduletlinktrate

ijij







11



(1)

where

)(trate

is a speed of data transfer from

microsatellite i to ground station

 j,

)(tlink

is a

mutual data transfer efficiency from microsatellite to

ground station,

)(tschedule

 – duration of data

transfer between the i-th microsatellite and j-th

ground station, planned according to the intervals of

their mutual visibility.

Therefore, a solution of the above problem

requires coordinated real time scheduling of tasks in

the stations and microsatellites network.

3 MULTI-AGENT TECHNOLOGY

FOR PROBLEM SOLVING

Traditional centralized planning is based on classical

mathematical methods, for example, linear and

dynamic programming, discrete optimization,

constraint programming, etc.

But it is well-known that such type of problems

is NP-hard and that is why a number of a new

heuristic and metaheuristic scheduling models,

methods and algorithms appeared, such as greedy

methods and other various methods of local

optimization, neural networks and genetic

algorithms as well as other evolutional

computations, simulated annealing, tabu search, ant

search, mixed miscellaneous metaheuristics and

many others (Leung, J. YT. (ed .), 2004).

However, there is still a gap with the market

requirements because all orders and resources need

to be known in advance and processed in batch

mode which it is not a case in real life.

Multi-agent technologies allow to solve complex

problems evolutionary on the fly, according to new

unpredictable events coming in a real time

(Wooldridge, 2009, Rzevski, Skobelev, 2014).



In developed approach,system will receive orders

in real time as well as a flow of other unpredictable

events: order cancellations, unavailability of

resources, failures or delays etc.

The plan for resource usage has to be

dynamically formed and continuously and

adaptively revised taking into consideration

individual set of criteria, characteristics, preferences

and constraints of orders and resources. The full

ECTA2014-InternationalConferenceonEvolutionaryComputationTheoryandApplications

154

cycle of satellites and stations management must

include fast reaction to new events, allocation of

orders to resources, scheduling of orders/resources,

optimization of orders (if time is available),

communication with users, monitoring of plan

execution, re-scheduling in case of a growing gap

between the plan and reality.

The revision of the schedule must be made by

the allocation of operations to open time slots or by

solving conflicts between operations that can be

shifted to previously allocated resource or re-

allocated / swapped to the new resources.

Communication with users means supporting a

dialogue with the users via mobile phones or other

tools initiated by either side at any time.

The developed approach is based on a “holon”

concept of PROSA system (Brussel, 1998) where

specific classes of agents of “orders”, “products”

and “resources” were introduced as well as a “staff”

agent which monitors results and advises other

agents when required.

To make this approach more flexible and

efficient the concept of Demand-Supply Networks

(DSN) was introduced where agents of demands and

supply are competing and cooperating on Virtual

Market (VM). In the concept any agent (holon) of

physical or abstract entity can generate “small”

demand and supply agents, which follow the specific

requirements.

As a result, the schedule can be formed as a kind

of requirement-driven network of operations which

can be easily adapted by events in real time

(Skobelev, Vittikh, 2003, Skobelev, Vittikh, 2009).

The core part of the method of adaptive

scheduling can be identified as the following:

1. The number of classes of demand and supply

agents represents specifics of the problem

domain with the required level of granularity.

2. Satisfaction function and function of bonuses /

penalties are represented by linear combination

of multi-criteria objectives, preferences and

constraints of each agent.

3. Protocols are defined which specify how to

identify conflicts and find trade-offs with the

open slots, shifts and swaps of operations.

4. A schedule formed in the process of DSN

agents self-organization is based on decision-

making and interaction of agents.

5. Special event procession protocols are

triggered when new events occur (for example,

arrival of a new demand):

a. An agent is allocated to a demand as it

arrives into the system. The Demand

Agent sends a message to all agents

assigned to available resources stating

that it requires a resource with particular

features and it can pay for this resource

with a certain amount of virtual money.

b. All agents representing resources with

all or some specified features and with

the cost smaller or equal to the specified

amount of money, offer them to the

Demand Agent.

c. The Demand Agent selects the most

appropriate free resource from those on

offer. If no suitable resource is free, the

Demand Agent attempts to obtain a

resource, which has already been linked

to another demand, by offering to that

demand some compensation.

d. The Demand Agent who has been

offered some compensation considers

the offer. It accepts the offer only if the

compensation enables it to obtain a

different satisfactory resource and at the

same time increase the overall value of

the system.

e. If the Demand Agent accepts the offer,

it reorganises the previously established

relationship between that demand and

resource and search for a new

relationship with resource increasing the

overall value of the system.

f. The same process is running for

Resource agents which are able to

generate Supply agents with specific

context-based requirements.

6. The above process is repeated until all

resources are linked to orders and there is no

way for agents to improve their current state or

until the time available is exhausted.

To achieve the best possible results agents use

the virtual money that regulates their behaviour. The

amount of virtual money can be increased by getting

bonuses or decreased by penalties depending of their

individual cost functions. The key rule of the

designed VM is that any agent that is searching for a

new better position in the schedule must compensate

losses to other agents that change their allocations to

resources, and propagation of such wave of changes

is limited by virtual money (Skobelev, Vittikh,

2009).

Therefore, the final schedule is built as a

dynamic balance of interests (consensus) of satellites

and stations agents that negotiate for their position in

the network schedule and plan their work by shifting

and reallocating time slots with the view on their

objectives and interest of the whole swarm.

RealTimeSchedulingofDataTransmissionSessionsinaMicrosatellitesSwarmandGroundStationsNetworkBasedon

Multi-AgentTechnology

155

4 MULTI-AGENT SYSTEM FOR

SESSIONS SCHEDULING

4.1 Architecture of the System

Architecture of multi-agent system for

communication sessions scheduling between

microsatellites and ground stations network is sown

in the Figure 1.



Figure 1: Architecture of multi-agent system for

communication sessions scheduling.

The key components of the system include:

• Physical Model – a simulation component that

is designed for Earth rotation simulation,

calculation of satellites orbits and visibility

intervals for microsatellites and ground

stations.

• Adaptive scheduler – scheduling system,

responsible for the forming and adaptive

change of the data communication sessions

schedule. Adaptive scheduler is based on the

multi-agent technology and is implemented in

Java Agent Development Framework (JADE)

according to FIPA standards for intellectual

agents (FIPA, 2014).

• Ontology provider – a component that allows

to formalize problem domain knowledge by

defining concepts and relation of the domain

area represented in a form of semantic

network;

• User Interface – provide possibility to make

settings and visualize results of scheduling.

All components are developed in Java.

4.2 Ontology of Small Satellites and

Ground Stations

An ontology is used to formalize the knowledge that

is necessary for the agents to take decisions.

According to our approach, knowledge should be

separated from the program code of the system and

kept in the ontology that represents a network of

concepts and relations of the problem domain area

(Huhns, 1997, Skobelev, 2012).

More specifically ontology in the developed

approach hels to specify requirements for satellities

and stations to make matching. Fragment of

ontology of small satellities and networks is shown

in the Figure 2 where it mainly includes

microsatellite, ground station and tasks scheduled

for the system.

Ontology helps to configure parameters of tasks,

microsatellites and ground stations and set the rules

of planning sessions between microsatellites and

ground stations.

Station

has

Schedule

Microsatellite

Сommunication session

expects

consists of (whole-part)

Priority Duration

Time interval

consists of (whole-part)

Start time End Time

visible in



Figure 2: Ontology fragment of the multi-agent system for

communication sessions scheduling.

According to this structure, each ground station

has individual schedule where planned

communication sessions with microsatellites and

station operation schedule are displayed. Each

communication session is characterized by priority,

duration, performance status and time period when it

need to be executed. Communication sessions can be

possible only in certain time intervals, during which

a direct visibility between a microsatellite and a

station is maintained.

4.3 Agents Functionality

Each ground station is associated with a resource

agent. Resource agent objective is to maximize

profit and schedule maximum of sessions with the

preference of own microsatellite tasks and providing

minimum idle run of the equipment.

Each communication session is associated with a

task agent. Task agent objective is to allocate to the

best ground station, which parameters fully satisfy

problem constraints with minimum costs. Task agent

can react to the events, addition/removal of ground

stations and tasks cancellations.

Satellite agent create session agents and is trying



Adaptive

scheduler

Ontology

provider

Physical

model

Ontology

User

Interface

ECTA2014-InternationalConferenceonEvolutionaryComputationTheoryandApplications

156

to maximize its own efficiency – to provide as much

useful information as possible.

To provide the economy-driven reasoning,

virtual market is introduced, where all decisions of

agents are linked with virtual money. Each agent has

an objective and bonus/penalty function.

The objective function represents satisfaction of

the agent depending on the achievement of the

objectives. Bonus/Penalty function sets a bonus for

achieving ideal target objectives or a penalty for

missing targets.

Each agent tries to maximize its satisfaction and

increases the profit (virtual money) that he can

spend for shifting other agents in the case of a

conflict, which he tries to resolve in the interests of

the system as a whole. Objective function of the

system is calculated as the sum of satisfaction of all

its existing agents (Figure 3(a)). Scheduling

problems is solved iteratively in evolution way by

the step-by-step increase (local improvement) of the

objective functions values of each agent.

When the task is initialized its agent gets the

initial sum of virtual money according to its priority.

This money can be used as a payment for allocation

on the chosen resource and for the compensation of

the expenses of the tasks shifted during the

allocation. Objective and penalty functions are built

considering time limits for the task agent and

considering the loading on the certain scheduling

interval fort the resource agent. T – time horizon,

KPI – key performance indicator (Figure 3 (b)).

KPI, %

T, hour



a) b)

Figure 3: An example of the task (a) and resource (b)

objective function.

The key rule of the designed VM is that any

agent that is searching for a new better position in

the schedule must compensate losses to other agents

that change their allocations to resources, and

propagation of such wave of changes is limited by

virtual money.

4.4 Agent Interaction Protocols

Scheduling process includes two phases which

continuously repeated:

1) Initial allocation of tasks to the resources

considering both preferences and time constraints;

2) Proactive improvement of tasks and

resources by their rescheduling.

Agent interaction protocol for the first phase is

shown in the Figure 4. This protocol is a

modification of Contract Net interaction protocol

implementation, specified by the FIPA standard.

Arrived task agent defines a list of available and

appropriate for the allocation resource agents and

then it sends a message-request CFP (Call for

Proposal) to each of the, which contains satellite

name and time interval, during which a

communication session should be scheduled.

Figure 4: First phase of agents interaction protocol in case

of a new task arrival.

Each resource agent that received this message,

forms a visibility intervals analysis in the specified

time slot. The calculations are only performed on

request, because visibility intervals calculation

between microsatellite and a station is a resourceful

operation. Then from the obtained list of the possible

task allocation intervals those time slots are

excluded during which station is busy according to

its schedule. If there are no time slots available for

allocation, resource agent sends a response “refuse”,

otherwise – message “propose” that contains a list of

the available intervals.

When resource agent receives an “accept-

proposal” message, it checks the specified

scheduling time slot for occupation by other tasks. If

the interval is free then a responding message

“inform” is sent, resource agents gets a payment for

the allocation and task information is recorded in the

schedule. Otherwise, a “failure” message is sent and

Task Agent

Resource Agent

CFP (satellite name, permissible slots )

refuse

i ≤ n

propose (list of available slots)

[there are no available slots for placing]

j = n - i

reject-proposal

j-1

accept-proposal (slot for placing)

failure

inform

[selected interval is not avai lable]

[task is scheduled]

Selection slot

RealTimeSchedulingofDataTransmissionSessionsinaMicrosatellitesSwarmandGroundStationsNetworkBasedon

Multi-AgentTechnology

157

when task agent receives it, it tries to be allocated on

the other resource.

Then an phase of proactive improvement of task

agent satisfaction is initiated. Task agent with the

smallest objective function value starts the

improvement process first. Proactive task asks the

appropriate resources, defining the allocation cost in

the different time slots.

Among the tasks already planned on this

resource, two tasks one on the left and one right side

of the slot are selected. Agents of these tasks receive

a request on shift on the specified time. Recursive

shifting of the tasks affected by the shift continues

until one of the tasks can move to the new position

without conflicts, the displacing task still have

means to compensate the expenses or a counter that

limits recursion depth equals zero. A process of

agent interaction when shifting the tasks in the

ground station schedule is shown in Figure 5 where

solid arrows represent messages with the shift

request and response messages of the shifted tasks

are shown as dotted lines.

Figure 5: Recursive task shifting.

From the set of possible allocation positions

those options are excluded, which confirmation will

not let to improve the system objective function

value, and the best option is chosen from the ones

that are left.

The task that remains unscheduled is out in the

list of tasks that wait for the scheduling. New

attempt of scheduling for these tasks will be made in

case of arising new events of adding new resources

or schedule changes of the existing ones.

The designed system implements models,

methods and algorithms that were earlier developed

for the multi-agent swarm of satellites (Sollogub,

2013), where the incoming tasks adaptively

reschedule tasks of satellites.

4.5 Example of Scheduling

Let us consider an example of communication

sessions scheduling for 5 ground stations.

In the traditional approach the data from

satellites will be transmitted only to their own

ground stations (Table 1). In these case the KPI of

all the involved stations is defined as the ratio

between the total data receipt time (~2720 sec) and

the length of the considered scheduling interval (24

hours) and it will be equal about 3%.

Table 1: Tasks allocation on its own stations.

Satellite Ground

Station

Sessions

number

Total

connection

time, sec

Data

volume,

CubeSat XI Kashima11 11 2200 2.64

HamSat Neustrelitez 17 3400 4.08

Mozhayets Kourou11 12 2400 2.88

ECHO Wallops14 15 3000 3.6

CubeSat V Malindi 13 2600 3.12

Total 16.32

With the use of multi-agent scheduling system

for the the same number of tasks it will be enough to

use only two stations. It allows not only to reduce

the maintenance cost of ground infrastructure and

increase stations KPI up to 20%, but also increases

the volume of transferred data from 16,32 to 20,64

Mb (with the data transfer speed 9600 bit/s).

Table 2: Communication sessions scheduling with stations

of the network.

Satellite Ground

Station

Sessions

number

Total

connection

time, sec

Data

volume,

CubeSat XI

Kashima11

Neustrelitez

11 3000 3.6

HamSat 17 4600 5.52

Mozhayets 12 2800 3.36

ECHO 15 3600 4.32

CubeSat V 13 3200 3.84

Total 20.64

Volumeincrease of the data transferred from the

satellite is shown in the Figure 6 that is achieved by

communication sessions scheduling in the network

of ground stations, comparing to the traditional

approach.

The results is showing the high value which

could be provided by bio-inspired models and

methods for real time scheduling based on

fundamental principles of self-organization and

evolution.

Figure 6: Transferred data volume.

ECTA2014-InternationalConferenceonEvolutionaryComputationTheoryandApplications

158

Next steps in the R&D work will be focused on

industrial version of the solution for modeling EU

Cubsat50 program and other applications.

5 CONCLUSIONS

The papers presents developed approach for

adaptive scheduling of data transmission sessions for

group of small satellites and ground stations.

It is shown that solution of the considered

problem evolutionary emerges from interaction and

trade-offs of many agents which continuously self-

organize themselves and change decisions to

improve their objectives and the objectives of the

system as a whole.

The advantages of developed multi-agent

solution are high adaptability, flexibility and

efficiency of future satellite services.

REFERENCES

Zinchenko, O., 2011. Small optical satellites ERS. –

[online] Available at: http://www.racurs.ru/?page=710

[5 April 2014].

Global Navigation Satellite Systems. United Nations. New

York, 2012. – [online] Available at: http://

www.oosa.unvienna.org/pdf/icg/2013/Ed_GNSS_eBo

ok.pdf [1 April 2014].

Leung, J. Y-T. (ed.), 2004. Handbook of Scheduling:

Algorithms, Models and Performance

Analysis,,Chapman & Hall / CRC Computer and

Information Science Series. London-New York-

Washington, 1st edition.

Wooldridge, M., 2009. An Introduction to Multiagent

Systems, John Wiley&Sons. London, 2nd edition.

Rzevski, G., Skobelev, P., 2014. Managing complexity,

WIT Press. London-Boston, 1st edition.

Brussel, H.V., Wyns, J., Valckenaers, P., Bongaerts, L.,

1998. “Reference architecture for holonic

manufacturing systems: PROSA”, Computer in

Industry, vol. 37, no. 3, pp. 255-274.

Skobelev, P., Vittikh, V., 2003. “Models of Self-

organization for Designing Demand-Supply

Networks”, Automation and Remote Control, Journal

of Russian Academy of Science, no. 1, pp. 177-185.

Skobelev, P., Vittikh, V., 2009. “The compensation

method of agents interactions for real time resource

allocation”, Avtometriya, Journal of Siberian Branch

of Russian Academy of Science, no. 2, pp. 78-87.

Foundation for Intelligent Physical Agents (FIPA) –

[online] Available at: http://www.fipa.org [25 March

2014].

Huhns, M., Singh, M., 1997. “Ontologies for Agents”,

IEEE Internet Computing, vol. 1, no. 6, pp. 17-24.

Skobelev, P., 2012. “Activity Ontology for Situational

Management of Enterprises in Real Time”, Ontology

of Designing, vol. 1, no. 3, pp. 6-38.

Sollogub A., Skobelev P., Simonova E., Tzarev A.,

Stepanov M., Jilyaev A., 2013. “Intelligent System for

Distributed Problem Solving in Cluster of Small

Satellites for Earth Remote Sensing”, Information and

Control Systems, vol. 1, no. 62, pp. 16-26.

RealTimeSchedulingofDataTransmissionSessionsinaMicrosatellitesSwarmandGroundStationsNetworkBasedon

Multi-AgentTechnology

159