Design of an Autonomous Decision Support System for High-Level Planning in Nano Satellites Using Logic Programming

(1)

Design of an Autonomous Decision

Support System for High-Level Planning in Nano Satellites Using Logic Programming

Saliha Serdar

Space Engineering, masters level 2017

Luleå University of Technology

Department of Computer Science, Electrical and Space Engineering

(2)

Design of an Autonomous Decision Support System for High-Level Planning in Nano Satellites Using

Logic Programming

Master Thesis in the course of the study programme

"Master in Space Science and Technology" by

Saliha Serdar

born on April 24^th 1991 in Groß-Gerau

Submitted on:

October 11^th 2016

Julius-Maximillians-University Luleå Tekniska Universitet

Department of Computer Science Department of Computer Science Aerospace Information Technology Electrical and Space Engineering

Prof. Dr.-Ing. Hakan Kayal Prof. Dr.Eng. Reza Emami

Prof. Dr. Dietmar Seipel

(3)

Statutory declaration

I confirm that this Master’s thesis is my own work and I have documented all sources and material used. This thesis was not previously presented to another examination board and has not been published.

Würzburg, 11.10.2016

(4)

Abstract

Low-level decisions in space missions, like maximizing the contact duration or bringing the spacecraft in safe mode in case of anomalies, are autonomously made by the spacecraft, whereas high-level and critical decisions are still taken by humans. Due to communication delays in interplanetary or even interstellar missions, this leads to the limitation of spacecraft operations in case of unexpected situations. Unexpected situations can be either the detection of unforeseeable short lived events or even on-board failures. In this given conditions the spacecraft have to take quick decisions to not miss the event or loss the spacecraft. Higher demands are imposed to spacecraft autonomy, if an event is detected and an on-board failure occurs at the same time. The presented work deals exactly with the last stated problem, which requires autonomy in high-level planning. A decision should be taken between either investigating the event or repairing the failure. Thereby the unique scientific measurements, that can result from the detected event, as well as the impact of the failure are considered. In order to reach this objective an approach of rule-based decision support system, also referred to as a expert system, is designed for nano satellites. For this purpose, events and on-board failures are defined, analyzed and converted from objective ratings into numerical values by applying the Analytical Hierarchy Process. Since the logical programming language Prolog is an appropriate language for experts systems, a part of the developed system is implemented in Prolog, to verify its use in space related expert systems.

(8)

Acknowledgment

First of all I want to thank my master thesis advisors Prof. Dr.-Ing. Hakan Kayal and Prof. Dr.

Dietmar Seipel of the department of computer Science at the University Würzburg. Prof. Kayal supported me during my thesis with his expert knowledge concerning aerospace technology and Prof. Seipel, as a Prolog expert, introduced me in Prolog. I would also like to thank Florian Kempf (research assistant at the University Würzburg) for inspiring me with new ideas, that helped me to make great progresses in my thesis.

Finally, I must express my very profound gratitude to my parents, to my partner and to my friends for providing me with unfailing support and continuous encouragement throughout my years of study and through the process of researching and writing this thesis. This accomplishment would not have been possible without them. Thank you.

Saliha Serdar

(9)

Acronyms

ADCS Attitude Determination and Control System ADIA++ Autonomous Diagnostic System for nano satllites

AEGIS Autonomous Exploration for Gathering Increased Science AHP Analytical Hierarchy Process

ANP Analytical Network Process

ASAP Autonomous Sensor And Planning ASE Autonomous Sciencecraft Experiment

ASTIA Autonomous Science Target Identification and Acquisition

CASPER Continuous Activity Scheduling, Planning, Execution and Re-planning ChemCam Chemistry and Camera

CI Consistency Index

DSS Decision Support System EDAC Error Detection and Correction EO-1 Earth Observing-1

ESA European Space Agency ESD Electrostatic Discharge

FDIR Fault Detection Isolation and Recovery FIDO Field Integrated Design and Operations

GESTALT Gird-based Estimation of Surface Traversability Applied to Local Terrain

GRB Gamma Ray Bursts

(10)

Acronyms vii

HG High Gain

HMNAO Her Majesty’s Nautical Almanac Office

HW Hardware

JPL Jet Propulsion Laboratory

KS Knowledge System

KSTIS Knowledge based Science Target Identification System

LG Low Gain

LIBS Laser Induced Breakdown Spectrometer

LS Language System

𝜇ASC micro Advanced Stellar Compass MBU Multiple Bit Upset

MEL Mars Exploration Laboratory MER Mars Exploration Rover

MISUS Multi-Rover Integrated Science Understanding System NASA National Aeronautics and Space Administration

OASIS On-Board Autonomous Science Investigation System for Opportunistic Rover Science

OBC On-Board Computer

OBSW On-Board Software

PCDU Power Control and Distribution Unit PPS Problem-Processing System

PROBA Project for On-Board Autonomy PROLOG Programming in Logic

PS Presentation System RCS Reaction Control System

RI Random Index

(11)

Acronyms viii

RIA Rock Identification Agent RMI Remote Micro Imager SEB Single Event Burnout SEE Single Event Effects SEL Single Event Latch-up SEU Single Event Upset SSTV Slow Scan Television

STFC Science & Technology Facilities Council

SV Science Values

TDL Task Description Language TID Total Ionizing Dose

TOMS-EP Total Ozone Mapping Spectrometer in NASA’s Earth Probe series TT&C Telemetry Tracking and Command System

USNO United States Naval Observatory

(12)

1 Introduction

Intelligent systems are becoming more and more a part of our daily life. Examples therefore are the digital assistances (e.g. Siri and Amazon Echo), autonomously driving cars (e.g.

Google Chauffeur ), computer games (to create challenges for the player), medical diagnosis systems (MYCIN [1]) and much more. But what exactly is the definition of intelligent systems? According to Gudwing (2000) [2] intelligent systems have the ability to work in a changing environment. Also in the space area intelligent systems are getting meaningful, but require a certain degree of autonomy. In a common mission, commands are uploaded to the spacecraft during the contact time window by the ground station. Afterwards they are executed sequentially by the spacecraft at a predefined time. Until the next contact, the spacecraft operates blind according to the uploaded commands. In case of unexpected situations, the spacecraft is not able to reschedule the commands in order to respond to changes. This can lead to significant drawbacks, if an unexpected event, which might be interesting to investigate, is missed by the spacecraft. Another difficulty is given regarding to the health status of the spacecraft. Failures and anomalies can be monitored by the ground station only during contact time. Of course the spacecraft is not totally alone with its failures and anomalies, there is a system called Fault Detection Isolation and Recovery (FDIR) on-board the spacecraft. As the name suggests, FDIR has the task to detect, isolate and recover the occurring failures.

However the isolation and recovery parts are extremely limited to only a few operations, like power down of the affected component, releasing the redundant element if the operating one failed or as the last invention change the state of spacecraft to safe mode [3].

With increasing distances between spacecraft and ground station, the stated operational limitations of spacecrafts are also increasing. For example a one way contact duration between mars rovers and ground stations takes approximately 20 minutes. Due to this fact teleoperation of mars rovers are impossible to realize. Since in case of an unexpected situations, e.g. slipping of the rover, there are no possibilities given to react in real-time. This is overcome with the supervised autonomy, where the destination is transmitted by the ground station and the rover decides autonomously about the interim goals. Some degree of autonomy is as well given in satellite missions, e.g. in NASA’s EO-1 mission, where the spacecraft is able to respond to unexpected events (2.5) and in ESA’s PROBA mission, in which the low level autonomy like pointing the camera to the desired position (2.6) are available. However the EO-1 spacecraft is

(13)

1 Introduction 2

a medium sized satellite with a mass of 572kg, which leads to high costs in development as well in launch. The satellites of the PROBA mission are small satellites with a mass range of 100kg up to 300kg, but still expensive and deliver a low level of autonomy.

Currently the department of Computer Science - Chair VIII of the University Würzburg is developing SONATE, a nano satellite which will be able to detect unexpected events and reschedule the command plan in order to investigate them. Additionally it will have the ability to diagnose its own health status. Detecting events and rescheduling the commands are the tasks of the payload ASAP, whereas the fault diagnosis will be done by ADIA++. Both payloads will operate autonomously, without an intervention from Earth. This project is funded by the German Federal Ministry of Economy Affairs and Energy, represented by the German Space Agency [4].

In the presented work a system, named Èxypnos System (éxypnos comes from the Greek and means intelligent), for high-level planning is designed. It will assist the spacecraft in critical decision making situations, which will increase the degree of autonomy. Here the critical situations are delimited by the occurrence of on-board failures and simultaneous detection of unexpected events. Thereby the decision have to be taken between either to apply a corrective measure to repair the failure or to investigate the detected event. The system is designed based on an invented nano satellite, called ÈxypnosSat, which is inspired by SONATE.

The designed system is an outline of an autonomous decision support system (DSS) for the above specified circumstances. Since the designed DSS will act like a domain expert, such systems are also called expert systems. For this objective the logical programming language Prolog is chosen due to its declarative proceeding, which suits well in expert systems. The focus of this work is placed to the analysis of on-board failures and unexpected events. Failures and unexpected events are converted from objective ratings into numerical values according to their degree of impact and importance respectively. Therefor the multi-criteria decision making approach Analytical Hierarchy Process (AHP) is applied. Based on these analyses an illustrative example of the power subsystem is implemented in Prolog to verify its use as well in space related expert systems.

The structure of the thesis is carried out as follows: As a first step a brief overview of the state of the art of autonomous and intelligent systems in the space area will be given in Section 2.

Afterwards in Section 3, the theoretical background of DSS, Prolog and the applied decision making approach, AHP will be declared. In Section 4 the design of the invented ÈxypnosSat will be outlined followed by its failure analysis in 5 and the analysis of unexpected events in 6.

After the failure and event analyses, the DSS will be designed in Section 7 and implemented in the logical programming language Prolog. Finally the results and future works will be discussed in Section 8 and in Section 9 the conclusion of the done work will be drawn.

(14)

2 State of the Art

Before designing the intelligent decision support system a research of already existing intelligent systems in space is made and presented in this chapter. There is no differentiation made between rovers and spacecrafts. Since the field of high autonomous spacecrafts is limited, the size and mass ranges of the investigated rovers and spacecrafts are as well not specified. In Section 2.1 - 2.4 intelligent systems in rovers will be addressed. Autonomous satellites will be stated in 2.5 and 2.6. After the state of the art of intelligent systems in space are outlined, a summarized review will be given in the Section 2.7.

2.1 On-Board Autonomous Science Investigation System for Opportunistic Rover Science - OASIS

Increased traveled distance of planetary rovers can increased the chance to gain high qualitative scientific knowledge. While NASA’s first successful Mars rover, Sojourner, covered a distance about 100m in the whole life time, one of NASA’s Mars Exploration Rovers, Opportunity, covered up to date about 43km. This major step forward in rover missions was realized with the autonomous driver software GESTALT (Gird-based Estimation of Surface Traversability Applied to Local Terrain). It provides the rover the ability to drive autonomously through the Martian surface to the desired destination. One problem here is, that with increased traveled distance the transmission time slots between Earth an Mars remain constant and are used in most cases for decision making purposes (e.g. detecting a rock of scientific interest is done by the ground control system). The consequent of this procedure is that in a long journey of the rover, most of the traversed terrains remain undiscovered [5].

In order to use the limited transmission time slots meaningful by transmitting more scientific data instead of commanding the rover, the OASIS system was developed by the engineers of NASA’s Jet Propulsion Laboratory (JPL). OASIS is able to recognize and analyze autonomously targets and events of scientific interest on-board the rover. Terrain features and events which requires further investigation can be directly identified by the rover. This system was tested

(15)

2.2 Autonomous Exploration for Gathering Increased Science - AEGIS 4

successfully by the FIDO¹ rover [5].

The OASIS system first detects predefined features based on the image data. These features are predefined by the scientific team members of the mission. After detecting features there are two ways possible for the further actions. Either an image segmentation can be done to categorize the sky and rocks followed by the extraction of the features or the characteristics are extracted directly from the input image. If this is done, the features, e.g. of rocks, will be analyzed and afterwards prioritized to define new scientific goals in case of interesting observation. Four different options are given to determine the target of scientific interest:

- Detected Event: sets flag if an event of interest is captured

- Key Target Signature: recognizes properties, that are predefined by scientists - Novelty Detection: recognizes properties with high deviation from usual values - Representative Sampling: identifies rocks that are representative for the traveled

region to gain characteristics of this region

OASIS has also the ability to reschedule the command sequence when an interesting feature is detected, to monitor the actual state of the rover and to execute the rescheduled commands.

Rescheduling of commands and monitoring rovers actual state is provided by the CASPER² system [6]. The execution of the commands are performed by the system, called TDL ³ [6].

Both systems, CASPER and TDL are integrated in OASIS.

2.2 Autonomous Exploration for Gathering Increased Science - AEGIS

AEGIS is a software, which is also developed by NASA’s JPL for planetary rovers. It is a part of the OASIS framework and allows the rovers to determine autonomously targets of scientific interest, in order to point the remote-sensing instruments. With AEGIS it is possible to increase the efficiency of the mission. Since a common target selection by scientist on Earth can take several days due to the transmission delay and during this time the rover has to stand at the same position for several days. The target selection with AEGIS is done on the basis of predefined criteria and constraints by human experts [7], that are uploaded to the rover.

The strategy of this software in the first instance is to analyze images on-board, which are provided by the navigation cameras of the rover. The result of this analysis is identification of potential targets. Based on this analysis relevant targets are extracted and prioritized depending

1is a prototype rover on Earth for testing purposes

2Continuous Activity Scheduling, Planning, Execution and Re-planning

3Task Description Language

(16)

2.3 Autonomous Science Target Identification and Acquisition - ASTIA 5

on their features (e.g. shape, size and surface reflectance). The prioritization is done thereby with the weighted sums each detected characteristics. The rating values of the characteristics are predefined constrains integrated in the memory of the rover. The relevant target with the highest priority is then chosen as the most interesting goal for scientific investigation [7].

AEGIS was first uploaded to one of NASA’s Mars Exploration Rover (MER) Opportunity in December 2009 in order to select targets for the narrow field of view Panoramic Camera, called PanCam. It is used to gain high-resolution color images of Martian sky and surface [8] to obtain geological and physical properties of Mars˜citeestlin2012.

After quite some time, in July 2016, the AEGIS software was also uploaded to NASA’s Mars Exploration Laboratory (MEL) rover Curiosity. Here the software is as well used to select targets of scientific interest with the navigation camera, but it points an other remote-sensing instrument, the Laser Induced Breakdown Spectrometer (LIBS) and the Remote Micro Imager (RMI) of Chemistry and Camera (ChemCam) instrument. The challenge compared with Opportunity is to select fine-scaled targets in order to point LIBS and RMI, since the diameter of LIBS is 0.3mm-0.5mm and the field-of-view of RMI is 1.15^∘ [9].

2.3 Autonomous Science Target Identification and Acquisition - ASTIA

The European Space Agency (ESA) makes also first steps towards on-board autonomy with the intended ExoMars rover, which was planned to launch at first in 2018 and later changed to 2020 [10]. The British government agency, Science & Technology Facilities Council (STFC), developed an OASIS like system (2.1), called ASTIA. It will identify targets of scientific interests and analyze surface sample autonomously on-board.

To reach the on-board autonomy, the ASTIA system is made up of several components: the Rock Identification Agent (RIA), the Knowledge based Science Target Identification System (KSTIS), the 3D Vision Agent and the Arm Agent [11]. After images are taken, RIA identifies the rocks with their relative centroids. This is an important key feature for the further investigation with the 3D Vision Agent, where the 3D coordinates of the target are extracted by stereo vision methods [11]. To rank the recognized targets according to their geological importances the KSTIS software is involved [12]. It is a fuzzy knowledge based expert system, developed together with experts from the field of geology. With respect to rock features (structure, texture and composition), KSTIS classifies detected rocks with Mamdani’s fuzzy-set method. The output of KSTIS are Science Values (SV) for each detected target representing its importance [11], [12]. The Arm Agent makes it possible to collect samples with the intended manipulator

(17)

2.4 Multi-Rover Integrated Science Understanding System - MISUS 6

of ExoMars rover. The Arm Agent involves the inverse kinematics of the robotic arm, to reach the desired target for sampling purposes.

2.4 Multi-Rover Integrated Science Understanding System - MISUS

In planetary missions a cooperation between several roves would increase new scientific dis- coveries. These rovers must have the ability to communicate and cooperate with each other to accomplish the entire mission. NASA is developing such a system, named Multi-Rover Integrated Science Understanding System (MISUS), to fulfill the imposed requirements. The essential requirements are highly autonomous rovers, to reach a maximum efficiency of rover operations with minimizing the communication with the ground station for decision making purposes. As a consequence, the rovers have to take their own decisions on-board.

The ability of cooperations of multiple rovers will be provided by the MISUS software. It’s main functions will be data analysis and distributed planning and scheduling. Data analysis will involve a machine-learning module to identify interesting features and discover them with setting new scientific goals. With this module the rocks can be analyzed and clustered regarding to their geological features. After clustering the investigated rocks, they can be prioritized relating to their importances, equivalent to the OASIS system (2.1). The main difference between MISUS and OASIS is given in the distributed planning and scheduling module. Similar like in OASIS the CASPER software will reschedule the mission plan if an interesting event or feature is detected. However in MISUS the planning software is divided in central planner, where one global mission is generated for all rovers and distributed planner, where each rover has a specific mission plan. Both modules are controlled by the continuous planning software CASPER.

2.5 Autonomous Sciencecraft Experiment - ASE

Up to the recent past, spacecrafts were not able to take decisions autonomously on the basis of observations. Autonomy is an important feature for interplanetary and interstellar explorations, since phenomenas with a very short appearance period can be missed, due to the delayed command transmissions. The ASE software, developed by NASA, enables satellites to fulfill their missions completely autonomously. The autonomy involves to analyze scientific data and to plan the next steps of the observation [13]. To recognize unexpected events autonomously, the images are analyzed with respect to the differences of previous investigated images. Implemented

(18)

2.6 Project for On-Board Autonomy - PROBA 7

algorithms make it possible to detect events (e.g. melt of ice, lava flow) and to discover them.

In oder to reach this autonomy, ASE is divided in the following components [14]:

- On-board science algorithms: to analyze interesting events, features,

- Robust execution management software: to make it possible to execute plans, - CASPER software to reschedule mission plans.

Since 2003, the ASE software is uploaded to NASA’s first spacecraft of the New Millennium Program, Earth Observing-1 (EO-1) [14], which was launched in the year 2000 [15]. The aim of this mission is to design and test new space application technologies [16]. EO-1 has a total mass⁴ of 572kg [15] and is able to detect and discover dynamical events on Earth autonomously.

Events of scientific interests for this mission are thermal anomalies, clouds, flood scene and changed environment [16]. As a result of on-board autonomy the down-link data for decision making is decreased and the down-link of highest science data is increased [16].

2.6 Project for On-Board Autonomy - PROBA

ESA is also willing to develop spacecrafts with on-board autonomy, which is the intension of the Project for On-Board Autonomy (PROBA) mission that is a part of the Technological Demonstration Program. With PROBA the operation by the ground station should be minimized. Actual flying spacecrafts of this mission are PROBA-1, PROBA-2 and PROBA-V and planned mission for the end of the year 2018 is the PROBA-3[17].

The first satellite PROBA-1, launched in October, 2001, is an Earth observation satellite with the aim to test and demonstrate on-board autonomy[18]. The provided autonomy of PROBA-1 includes low level operations and resource management, camera pointing and scanning based on input data⁵, planning and execution of payload operations and communication with ground station[18].

PROBA-2 is the successor of PROBA-1 and was launched in November, 2009 [19]. The mission of PROBA-2 is Sun observation for space weather purposes. The autonomy of PROBA-1 is adopted and extended with an autonomous star tracker, named micro Advanced Stellar Compass (𝜇ASC).

The last realized PROBA mission, PROBA-V was launched in May, 2013 and is able to detect and differentiate autonomously land and sea[18]. This mission was also adopted and extended based on previous PROBA spacecrafts. The V in PROBA-V stands for vegetation and therefore the interesting areas are lands. A land-sea mask, a given map where lands and seas are marked,

4total mass is with propellant

5the input data are geographical coordinates, latitude and longitude

(19)

2.7 Conclusion of the State of the Art 8

makes it possible to differentiate between land and sea autonomously on-board. The estimation from the actual position up to the position 10 minutes in the future is possible. The camera switching is done autonomously by the spacecraft by means of the land-sea mask. The switching ON of the cameras can be done either by detecting land or by passing through a predefined geographical coordinate. As usual in spacecrafts, a Failure Detection, Isolation and Recovery (FDIR) system is also on board of PROBA-V. Once an anomaly or failure is detected by FDIR and the spacecraft is in the autonomous observation mode (called nominal mode), the following three possibilities for isolation and recovery are given:

- power cycle resource,

- switch to redundant resource,

- switch to system safe mode in case no redundant resource is available at that moment.

If it is possible to overcome the anomaly with the first or the second solution, then the spacecraft will stay still in the nominal observation mode.

The next planned spacecraft of the PROBA series is PROBA-3 and it will be the first step of the ESA towards formation flying. It is intended to launch two satellites in high elliptical orbits⁶ to fly them in precise formation with accurate pointing capability [20]. Acquired knowledge form previous PROBA mission will be deployed in this mission as well, especially the on-board autonomy.

2.7 Conclusion of the State of the Art

The research delivers the result, that both rovers and satellites have not the ability to handle autonomously in critical situation, e.g. an failure occurrence and event detection at the same time. Besides the autonomous navigation which is required in interplanetary missions, the autonomy of rovers are limited by target detection based on predefined features by experts. In case of on-board anomaly and detection of an event of scientific interest, the operators on Earth have to intervene. If e.g. a target is visible for a short time, a unique scientific measurement can be missed in this situation due to communication delay. The same problem is also given in EO-1 and PROBA satellites. Irrelevant what kind of strangeness the event has, e.g. the FDIR system of the satellites will change form observation mode into safe mode if the problem can not be fixed or the ground station have to interact with the satellites. Furthermore it is noticeable that intelligent systems are implemented up to now only in spacecrafts with high mass ranged from approximately 1000kg (e.g. Curiosity rover) to 100kg (PROBA-1). Spacecrafts with high masses are always coupled with high costs and therefore the mission is risk-aver.

6high elliptical orbit: low altitude perigee and high altitude apogee

(20)

2.7 Conclusion of the State of the Art 9

Based on this research, it can be stated, that the spacecraft autonomy in critical situations is an unexplored area. Concluded to this investigation an untouched field will be addressed by designing an intelligent system for nano satellites, that will support the spacecraft with a decision in case of critical situations. As stated before, a critical situation is specified by concurrently occur of failures and unexpected events. The basic concept of target selection by rovers, where the features are rated by values, is taken up and will be applied in the designed system.

(21)

3 Theory

In this chapter fundamentals will be presented and help to understand the designing process of the developed decision support system. In Section 3.1 the definition of an decision support system will be introduced firstly, followed by declaration of Prolog terms in Section 3.2. The theory of the applied multi criteria decision making approach, Analytical Hierarchy Process (AHP) is addressed in Section 3.3. This section involves the description of the used software Super Decision for the AHP method as well the reason why the AHP is preferred over the known simple scoring model.

3.1 Definition of Decision Support System - DSS

Decision making is a challenging task especially in complex systems. Furthermore a right decision making involves always an expert in the process. A system which supports and improves the judgment of decision makers and experts is provided by a so called decision support system (DSS). The problems involving a DSS, are usually unstructured or semi-structured, meaning that the problem can change rapidly its state and is not predictable [21]. A DSS is able to provide rapidly decision, when it is required in time critical problems. A specific definition of a Decision Support System is not given, that leads to not clearly defined characteristics [21].

According to BURSTEIN (2008) [22], the main components of a DSS are the language system (LS), the presentation system (PS), the knowledge system (KS) and the problem-processing system (PPS). The LS defines the commands, which can be translated by the DSS, whereas in PS the output vocabulary of the DSS is defined. The KS involves all informations about the problem stored partially in a database. The last listed component PPS is a problem solver component of a DSS.

Furthermore there exist several classifications of DSS frameworks like text-oriented, database- oriented, spreadsheet-oriented and still more, which can be found in [22]. For this work a rule-oriented DSS is intended. In a rule-oriented or rule-based DSS, the decision is taken based on predefined rules. These rules can be either extended by humans manually or in case of artificial neutral network, the system can define rules based on actions and results. If the rules

(22)

3.1 Definition of Decision Support System - DSS 11

are extended by the system itself, than the system is called a learning system.

A rule-based DSS is also categorized as expert system, since the experts knowledge is imitated in the rules [22]. This is used in case of the human expert is not available at the moment, if a time critical decision have to be taken [22]. An other factor for the absence of human experts are high costs, since a expert system can replace a human expert. The replacement underlines the difference between an expert system and a DSS, since in a DSS the expert is not replaced, but supported, whereas in expert system the expert is replaced.

Rule-Based Decision Support System - Expert Systems

The designed system in this work is a rule-based system and therefore a detailed definition of rule-based systems will be introduced. According to NEGNEVITSKY (2011) [1], the development of a rule-based system involves a domain expert, knowledge engineer, programmer and project manager. The domain expert is the person with a huge knowledge about the specific area gained by long-standing experiences. The knowledge of the human expert will be transferred to the expert system. The task of the knowledge engineer is to design and test the expert system based on the expertise of the human expert. His task involves also selecting the best programming language for the given problem. After this is done, a programmer with symbolic programming skills translates the knowledge in form of rules in a programming code.

And the last member, the project manager guides the whole team and is the interface to the users. It is possible to reduce the number of the development team with using expert system shells. Expert system shells are software for developing rule-based expert systems with less programming skills than required. The knowledge can then be directly defined as rules. With such softwares a small rule based expert system can be developed also only by one person [1].

As mentioned before the developed DSS in this work is a rule-based system or also called production system. A production system is based on "IF-THEN" clauses, also referred to condition and action clauses [1]. The condition is made up of at least one object and one value.

An example therefor is

IF ’traffic light’ is red

in which ’traffic light’ is the object and red the value. If the given object has the specified value then there is a consequence, called action. As well the action can be divided in two parts similar like in the condition part but does not require. It should be noticed, that the condition part requires at least one object and one value. The continuation of the above mentioned example for the action part is then

THEN stop.

(23)

3.2 Logical Programming Language - Prolog 12

3.2 Logical Programming Language - Prolog

To develop a rule-based DSS, a logical and symbolical programming language is required.

Prolog is the mostly used programming language for logic programming (Programming in logic). In this section a short introduction into Prolog is presented, where the essential Prolog terms will be introduced.

Prolog is a declarative language, that is made up of three components - facts, rules and queries.

Declarative programming languages are outlined with their abstract mode of expression of logical computations. Such languages enable domain experts to handle easier with the semantics of the program, since declarative languages do not focus on how a given problem has to be solved like imperative programming languages. They deal with the question what is the problem to be solve [23].

The user is able to ask the Prolog program question to solve the given problem of a specific domain. The posed questions to Prolog are called queries. With them it is possible to search through the facts and rules to deliver all correct and possible solutions. Prolog is a common used language in expert systems.

According to BRATKO (2001) [24], a Prolog program consists of clauses, where each of them ends with a full stop. Types of clauses can be distinguished by facts, rules and queries. Facts have the head form and consist of a functor with a defined arity. Arity is the number of arguments related to a functor. The arguments can be either atoms (constants) or variables (general objects). Examples of facts are

female(ann).

parent(ann, bob).

, in which the first fact has the arity 1, with the argument ann and the second fact has the arity 2 with the arguments tom and bob. The combination of a functor and arity is called predicate [25]. Predicates are either predefined by the Prolog system and called built-in predicates or are defined by the user as facts and rules, called user-defined predicates. The facts can be state as functor/arity, which are in the given examples female/1 and parent/2 [26]. The first fact is reading as "ann is female" and the second one "ann is parent of bob". These are user-defined predicates. One example of built-in predicate is the write/1 predicate,in which the argument of the functor write is given as an output on the console.

Rules are made up of the form head :- g_1, g_2, ..., g_n, in which head is the same head defined in facts, :- is the neck operator indicating the if clauses and g_1, g_2, ..., g_n is the body of the clauses consisting of n-goals [27], [25]. An example of a rule is

(24)

3.3 Analytic Hierarchy Process - AHP 13

mother(X, Y):-

parent( X, Y), female(X).

, in which the arguments in the functor are in this case variables. A variable in Prolog begins either with a capital letter or with an underscore character [27]. The then clauses of an if-then are written in Prolog after the head of the rule. An and clause in Prolog is defined by a comma.

The given exemplary rule is reading as, IF X is parent of Y and X is female, THEN X is mother of Y. Rules are stated as true if the goals predefined by facts are fulfilled, otherwise they are stated as false. A Prolog program can be extended by adding rules and facts without any problems.

After facts and rules are set, the user can ask the implemented Prolog program questions. The question must be typed after system prompt, which is a question mark followed by a hyphen

?-. The user does not need to type it manually, since Prolog generates it automatically on the console. A query is made up at least one goal, which has the same form as the facts. For the above introduced example of facts and rules, the question "is ann mother of bob?" can be asked with

?- mother(ann, bob).

, where the query ends with a full stop, since as mentioned before, it is also a clause. The rule defined above is applied and the answer of the Prolog system is true since the facts parent(ann, bob). and female(ann). are fulfilled. The variables X and Y are substituted by the atoms ann and bob respectively.

Up until now, a Prolog implemented decision support system is not used in space related missions. In NOGUEIRA (2001) [28] an A-Prolog decision support system is designed for the Reaction Control System (RCS) of Space Shuttle. RCS is relevant for maneuvering the spacecraft, while it is in space. It is computer controlled during take of and landing, whereas during the flight it is controlled by the astronauts. Since in critical situations the astronauts have to communicate with the ground station, an intelligent system implemented in RCS would be helpful. Such a system was designed successfully and conformed the use of the declarative programming language, but it was not being used in a real mission ([28]).

3.3 Analytic Hierarchy Process - AHP

There exist several types of decision theory techniques. The designed decision support systems are based on the Analytic Hierarchy Process (AHP). It is a concept for multi-criteria decision making and is developed by the mathematician Thomas L. Saaty [29]. With AHP it is possible to convert subjective evaluations into numerical values. Commonly this method is used in

(25)

3.3 Analytic Hierarchy Process - AHP 14

multi-criteria decisions, where applying AHP delivers the choice of the best alternative. Besides, AHP can be applied in wide range of decision making methods and one of them is the evaluation of the alternatives [29]. The AHP will be applied in the designed Èxypnos System to rate all possible failures and all possible events with a value.

SAATY (2012) describes in [30], that the easiest way to structure a decision problem is a three level hierarchy that consists of the goal of the decision, criteria and alternatives. Figure 3.1 depicts such a simple three level Hierarchy. The aim of a hierarchy is to consider by the decision also the elements in the level linked above.

The most challenging and creative part according to SAATY (2012), [30], is to define criteria in order to build the problem in a hierarchy. The criteria should consider the environment within the problem and the features influencing the problem. As illustrated in 3.1 the hierarchy does not have to be completed, it is possible that one element is not linked with all elements beneath, but at least with one. This not complete hierarchy exists, if the criteria are divided in sub-criteria and then linked to the alternatives.

The decision making process AHP is based on relative measurements [31], in which one criterion, for example A, is compared pairwise with an other criterion, B [30]. Here the pairwise comparison is only done for homogeneous elements. For the comparison the so called fundamental scale is used, which is also defined by Saaty, [30]. With these pairwise comparisons a square matrix for the criteria or sub-criteria is set up. Out of the square matrix the eigenvectors of the principal eigenvalue is calculated. The calculated eigenvector represents the weighting of each criterion or sub-criterion. This was only a rough overview of the AHP, a detailed description follows in the next subsection.

Figure 3.1: Three Level Hierarchy of the Analytic Hierarchy Process.

(26)

3.3.1 Detailed Approach of the Analytical Hierarchy Process 15

Table 3.1: The Fundamental Scale according to [30].

Intensity of Importance Definition Explanation

1 Equal importance Two activities contribute equally to the objective

2 Weak

3 Moderate

importance

Experience and judgment slightly favor one activity over

4 Moderate plus

5 Strong importance Experience and judgment strongly favor one activity over

6 Strong plus

7 Very strong An activity is favored very strongly over another; its dominance

demonstrated in practice

8 Very, very strong

9 Extreme importance The evidence favoring one activity over another is of the highest

possible order of affirmation

3.3.1 Detailed Approach of the Analytical Hierarchy Process

In this section the AHP will be explained step by step. An application of the method can be found in 5.2.2, in which AHP is applied to evaluate the power subsystem failures by numerical values.

Step 1. The first step is to divide the given decision problem into levels consisting of a goal, criteria, if appropriate sub-criteria and alternatives. As mentioned before this part is the most creative part to solve. The relationship between the levels is given with the connections to the above element, which is illustrated in 3.1. In case of classifying the criteria further into sub-criteria, there would be an additional level between criteria and level for sub-criteria. In this case the criteria will be linked to the sub-criteria and these in turn will be linked to the alternatives.

Step 2. The next step is to compare each criterion and if defined sub-criterion pairwise. This comparison has to be done for homogeneous elements. This means all criteria are compared with each other, whereas all sub-criteria related to one criterion are compared pairwise. Comparing sub-criteria across criterion is not given and does not make sense. The comparison is scored with the fundamental scale (3.1). In the most cases the pairwise comparison is done by experts or decision makers. It should be noticed that the pairwise comparison of the alternatives should

(27)

3.3.2 Super Decision Software 16

also be done with respect of the connected criteria or sub-criteria.

Step 3. Out of the pairwise comparison a square matrix, named comparison matrix, is set up, which diagonal entries are one. The other elements are based on the pairwise comparison.

Lets say i is the row of the matrix A and j the column. If the i^𝑡ℎ element is stronger than the j^𝑡ℎ, then the entry in the matrix A at the position (i, j) is larger than 1. The element at the position (j, i) is given by its reciprocal. But if the j^𝑡ℎ element is stronger than the i^𝑡ℎ element j, then entry at the position (i, j) is the reciprocal of the value, which states the importance of the element j based on the fundamental scale. And as well here the element at the position (j, i) is given by its inverse.

Step 4. The comparison matrix is build to derive the priority vector, w. This is done with the aid of eigenvector and eigenvalue method. The eigenvector of the principal eigenvalue is the priority vector w. How the eigenvalues ad eigenvector are derived will be not explained in this work but can be found in [31]. However by applying the AHP method a software (like Expert Choice or Super Decision) is usually used, in which eigenvalues and -vectors are derived.

Step 5. In order to check the consistency of the pairwise comparison done in step 2, the consistency ratio CR has to be calculated. It is the ratio of the consistency index CI and the random index RI. CI is given by

𝐶𝐼 = (𝜆_𝑚𝑎𝑥− 𝑛)

(𝑛 − 1) , (3.1)

in which 𝜆𝑚𝑎𝑥 is the maximum eigenvalue and n the order of the comparison matrix. RI is the average estimation of CI of randomly generated matrices and can be found in [31]. If the calculated CR is larger than 0.1 it exhibits the inconsistency of the pairwise comparison.

Step 6. In the last step all values of connected criteria, sub-criteria and alternative are multiplied, which provides the evaluation of each alternative respectively to the rating of the criteria and alternatives.

3.3.2 Super Decision Software

Due to the complexity of the Analytical Hierarchy Process, a software is necessary, which delivers the priority vectors described in previous subsection. In this work the Super Decision software is used. The hierarchic structure of the problem and their connections are done by the user himself, as well the pairwise comparison of homogeneous elements. The Super Decision software generates during the pairwise comparison the comparison matrices and calculates the related priority vectors with their inconsistencies. There is no requirement to derive the

(28)

eigenvector of the principal eigenvalues manually, which represents the priority vector. Since there are many matrix multiplication, it is useful to involve a software, which is either self implemented or already existing. There are several softwares for the AHP, but Super Decision is a free educational one. In this subsection a short introduction to the Super Decision software will be provided. A detailed tutorial of the Super Decision software can be found in [32].

The levels goal, criteria and alternatives are named in Super Decision software clusters. A cluster consists of elements, also called nodes. If a cluster is linked with a line to an other cluster, than the elements within the clusters are connected. It is possible to check which elements are connected by the Show Connections icon. The goal and criterion clusters can be named arbitrarily, whereas the alternatives cluster must involve the word "Alternatives".

Figure 3.2 illustrates a sample model of a car hierarchy, which can be loaded by the data name Ca_hierarchy.sdmod. E.g. the cluster 2Criteria consists of the four elements 1Prestige, 2Price, 3MPG and 4Comfort. All these elements are connected to the elements of the 3Alternatives cluster. As well the Goal Node element in the cluster 1Goal is linked to the elements of the 2Criteria cluster.

After all clusters and elements are build and linked, the pairwise comparison of elements within one cluster with respect to the connected element can be done. The pairwise comparison will be made for explained sample model Car_hierarchy. The pairwise comparison can either be done directly in the comparison matrix illustrated in 3.3 or in the so called questionnaire, which is depicted in 3.4. Both alternatives deliver the same result as it can be see in the figures on the right hand side in the part 3.Result. This is the priority vector for the done comparison, in which on the top the inconsistency is given. The same part is as well involved in the questionnaire comparison. In Figure 3.3 the blue colored values indicates the dominance of the elements on the left hand side, whereas the values written in red indicates the dominance of the elements listed on the top. During the pairwise comparison the priority vector is generated step by step. The inconsistency is increasing with increasing number of already done comparison.

This can help the user of the software to control the inconsistency and not exceed the value of 0,1. For the pairwise comparison the fundamental scale (3.1) is used. In the questionnaire if the element on the left hand side (blue) is more important than on the right hand side (red), than the scoring is done on the left scale. Inversely if the element on the right is more important, than the scoring have to be done on the right hand sided scale. Anyway which comparison method is chosen (matrix or questionnaire), as mentioned before both will supply the same priority vector and the same inconsistency.

If all pairwise comparisons of each element within a cluster are accomplished, the weighting of the alternative elements can be obtained. Therefore the Synthesize icon have to be selected in the software. A window will appear in the screen, which is depicted in 3.5. In this window the ratings off the defined alternatives are presented. For the design of an DSS only the columns

(29)

Figure 3.2: Shortcut of a Sample Model, Car Hierarchy, from Super Decision software.

Figure 3.3: Shortcut of Pairwise Comparison Window with Comparison Matrix.

(30)

3.3.3 Advantages of AHP over the Simple Scoring Model 19

Figure 3.4: Shortcut of a Pairwise Comparison Window with Questionnaire.

Normals and Ideals are of interest. The first one represents the priority vector mentioned in 3.3. The second one involves the normals values divided by the maximum Normals value. In this example the maximum Normals value is given by the alternative 3Honda Civic, thus the Ideals value leads to 1,0.

It should be noticed, that entire scores are given in percentages, both the priority vector resulting after the pairwise comparison and the priority vector of the alternatives (Normals).

As a result the Ideals are as well given in percentage. The purpose of Ideals is to rate the best alternative with 100,0%, but the proportions remain the same as in Normals. The analysis delivers in this case that the alternative 1Acura TL is 75,58% as good as the alternative 3Honda Civic and 2Toyota Camry is 43,95% as good as 3Honda Civic.

3.3.3 Advantages of AHP over the Simple Scoring Model

In this section a brief explanation will be given, why the AHP is preferred over the simple scoring model. With the simple scoring model, the intuitive scoring of criteria by experts and summing them up for the ranking of the alternatives, is meant.

The AHP approach for multi criteria decision making does not only involve the intuitive weighting of the given criteria, there are mathematically calculations behind it. Whereas the simple scoring model is based only on subjective judgments and basic mathematics (multiplying and summing). In both methods the ranking will be in the same order. For the purposes of the designed expert system not the ranking is of importance, but rather the rating of each alternatives. With AHP the evaluation of each alternative are preciser and more significant than in the simple scoring model. However due to pairwise comparisons the AHP approach is

(31)

3.3.3 Advantages of AHP over the Simple Scoring Model 20

Figure 3.5: The Scoring of the Alternatives of the Car_hierarchy Sample Model.

more time consuming than the simple scoring model. Furthermore the inconsistency factor, provided by AHP, method leads to overcome mismatches of the criteria ratings. Discrepancies of criteria ratings are given if e.g. the criterion A is more important than B and B is more important than C and C is more important than A.

Based on this advantages instead of applying the simple scoring model, the AHP is selected as the multi criteria decision making approach for the intended intelligent decision support system.

(32)

4 Spacecraft Mission Design

Before the rule based decision support system can be designed, a satellite mission has to be created. In this work the hypothetical space mission is invented and will be presented. The satellite of this mission has the name ÈxypnosSat, which is composed of Èxypnos (derives from the Greek and means intelligent) and satellite. The fictional ÈxypnosSat is based on SONATE, which is currently in development by the University of Würzburg and will be launch in 2019 [4]. It should be noticed that the design of the mission is simplified and not detailed. It serves the purpose to develop a decision support system for a nano satellite.

The invented ÈxypnosSat is a nano satellite for earth observation and has the aim to test and develop high-level on-board autonomy for future interplanetary or interstellar missions.

ÈxypnosSat must demonstrate the ability to detect and investigate not predictable events on and around Earth. If an anomaly of the spacecraft monitored and an event is detected at the same time, than the satellites have to decide between fixing the failure or investigating the event. Thereby the decision is influenced by the impact of the failure and the importance of the event.

Since it is a first step towards high-level autonomy, it is an earth observation mission. Greater benefits can be obtained in interplanetary and interstellar missions. Because in common missions the decision is taken by the operators on Earth and with increasing distance between spacecraft and ground station, the communication delay is also increasing. As a result unpredictable and short lived events will be missed, that maybe will never occur.

A short overview of the SONATE mission will be given in Section 4.1 and afterwards the design of ÈxypnosSat will be presented by firstly defining its orbit in 4.2 and then specifying the subsystems together with their related components in 4.3.

4.1 SONATE

Typically spacecrafts are controlled by the ground station. The spacecraft transmits to the ground station telemetry data and based on these the operators informs the spacecraft about the next steps via telecommand. Within the Earth orbit this leads to no complications. But in

(33)

4.2 Orbital Design 22

interplanetary missions, e.g. Mars mission, the communication between ground station and spacecraft will have a large delay due to the distance. This can lead to miss the not predictable event, with a short-time occurrence. This problem can be solved with an autonomy on-board the spacecraft.

The key mission of SONATE is to increase the on-board autonomy. This will be done by autonomously detecting not predictable events and rescheduling the command sequence to not miss the event. Furthermore it will be able to detect, analyze and forecast on-board anomalous that will occur in the future [4].

The nano satellite, SONATE, is been currently developing by the University of Würzburg. The operational lifetime of SONATE is set to one year and its aim is the in-orbit verification of the Autonomous Diagnosis System (ADIA) and the Autonomous Sensor and Planning (ASAP) system [4]. Both systems are described in Section 4.3.6. Further components for in-orbit verification are reaction wheels, AROS (4.3.3) and SSTV camera (4.3.6).

4.2 Orbital Design

The design of a spacecraft orbit does not offer any strict specifications, but for earth observation it is obvious to select as an orbit type the Earth-referenced orbit for Earth coverage [33]. Due to the fact that a polar orbit can cover the Earth nearly global [34], a polar orbit is chosen for ÉxypnosSat mission. The orbit of a spacecraft and its position is uniquely defined with the six Keplerian elements (also known as orbital elements). The meaning of each orbital element will be not declared in this section, but can be found in [33]. A typical polar orbit has an altitude of approximately 700km and an inclination of approximately 95^∘. Since for the first approach of the decision support system the elements are not required and therefore they will be not defined in this work.

4.3 Spacecraft Subsystems

More important than the orbit design for the decision support system are the subsystems of the spacecraft. Due to this fact, the subsystems will be explained in more detail. A spacecraft is divided in several subsystems and they are interdependent [35]. To have a fully functional satellite, each subsystem have to fulfill at least its purposes. The subsystems are differentiated between payload and satellite bus. The payload is individually specified for each spacecraft according the defined mission to fulfill it and therefore are the sole reason to get a satellite into space. The payload is not functional without the satellite bus, therefore its task is to enable

(34)

4.3.1 On-Board Computer - OBC 23

the payload to accomplish the mission and keep it healthy. In general a spacecraft’s satellite bus consists of six different parts:

1. On-Board Computer - OBC 2. Power System

3. Attitude Determination and Control System - ADCS 4. Thermal Control System

5. Telemetry Tracking and Command System - TT&C 6. Structure and Mechanism.

The structure and mechanism subsystem is not considered in this work for simplification purposes. In the following sections all other subsystems (1. - 5.) and the payload of ÉxypnosSat will be described in more detail with their related components (presented in Figure 6.1).The most critical and error-prone components of subsystems are redundant, in order to enable the spacecraft reaching the intended lifetime. In Figure 6.1 the number of redundant elements of the components is given in the brackets. In case of no brackets, non redundant element is available.

According to WERTZ (1999) [33], spacecraft redundancy can be categorized in either same design redundancy or functional redundancy. Same design redundancy is given if minimum two identical components exists and at least one of them is active. FORTESCUE (2011) [34] divides the same design redundancy in standby redundancy and active redundancy. In standby redundancy, the redundant element is turned off until the active element fails. In case of active redundancy all components are active and are sharing the load. If there occurs disagreements between active redundant elements, a voting process is applied. If there are no identical redundant elements but elements pursing the same aim, then a functional. One simple example for functional redundancy is the high gain and low gain antenna, since both are transmitting telemetry and receiving telecommand (but with different gains). It should be noticed that functional redundancies are not outlined in the figure 6.1. In the following subsections each subsystem will be presented.

4.3.1 On-Board Computer - OBC

The key subsystem, that controls the spacecraft is the on-board computer. It has a processing capability and is linked to all other subsystems through their components. The OBC runs the on-board software to enable the remote operations, to control functionalities and to monitor of the health status of the spacecraft. Moreover the OBC involves the components processors, memories and the software. The processors are the cores of OBC and are responsible for all calculations and algorithm implementations and as known from the usual memories on Earth,

(35)

4.3.2 Power System 24

Figure 4.1: Subsystems of ÉxypnosSat

the function of the memories in satellites is also to store data. It is an important component, since during the time in which no contact to the ground station can be established, all collected data are saved on the memories. Typically a spacecraft consists of more than one memory type [3]. The boot loader for the OBSW is stored in the boot memory, which is non-volatile ROM. The on-board software is stored in the work memory and the storage of the spacecraft’s health status takes place in the safeguard memory. Since the satellite has not permanently contact with the ground station to transmit telemetry and scientific measurement data, until a broadcast takes place these are stored in the science and housekeeping data memory [3].

4.3.2 Power System

The power system gives inanimate subsystem "life", since the main function of it is to provide the subsystems with energy. A common power system is composed of three main components - primary energy source, secondary energy source and Power Control and Distribution Unit (PCDU) [34]. The primary energy source in ÈxypnosSat mission is solar arrays. They are converting the gained solar energy into electrical power. During the sun light duration, the

(36)

4.3.3 Attitude Determination and Control System - ADCS 25

satellite uses the energy directly from the solar panels and charges the secondary energy source - the batteries. If the satellite is in eclipse duration, then the batteries will provide power to the subsystems. The PCDU decides about the switching between solar arrays and batteries, energy distribution to other subsystems and charging the batteries [34].

4.3.3 Attitude Determination and Control System - ADCS

It is important to know the position and orientation of the spacecraft, to orient, e.g. the payloads to the desired position to fulfill the mission or the solar arrays towards sun to gain energy. These requirements are met with the attitude determination and control system (ADCS). Sensors enable the orbit determination and actuators the orbit control, whereby a distinction between reference sensors and inertial sensors are made. References sensors measure the direction of the spacecraft relative to earth with reference points, like sun, stars or earth’s magnetic field lines, whereas inertial sensors measure only the change of spacecrafts attitude [34]. Therefore an inertial sensor have to collaborate at least with one reference sensor [34].

In ÈxypnosSat sun sensors, star sensors and magnetometers are used as reference sensors. Sun sensors are implemented to determine the direction of the sun in order to orient the solar arrays towards sun. Only sun sensors are not enough to determine the pose of the spacecraft.

Therefore additionally star sensors, magnetometers and gyroscopes are used. Star sensor can determine the pose of the spacecraft with high accuracy by using suitable star images and a star catalog. Usually star sensors have a high mass, big size and a high-level of energy consumption [34]. Therefor a star sensor, that suitable for nano satellites is required. Within the AROS project such star sensors are been currently developing by the University of Würzburg. The star tracker AROS is intended for ÉxypnosSat for precise attitude determination. Another type of reference sensors for attitude determination are the magnetometers. It provides both the magnitude and the direction of the magnetic filed relative to Earth. Indeed magnetometers are light and have a low power consumption but they are inaccurate.

For the invented mission only one inertial sensor type, the gyroscope, is intended. A gyroscope enables the measurement of spacecraft rotation starting from an initial start position. As described previously a gyroscope alone is not able to gain information about the position relative to Earth, hence it has to be combined with a reference sensor, e.g. magnetometer.

4.3.4 Thermal Control System

The components within the spacecraft can survive during the whole mission, if the required temperature intervals are not exceeded. The thermal control subsystem ensures, that the

(37)

4.3.5 Telemetry, Tracking and Command System - TT&C 26

temperature in the satellite is kept between these intervals. With respect to different subsystems, there is a distinction to be made between survival limits, which are always valid and operational limits, which are valid during operational mode [33]. The temperature in the spacecraft is measured with thermal control sensors. The temperature is maintained passive and active.

Passive thermal control is done by the design of the spacecraft, mechanical structure and materials (e.g. insulation) and does not need any kind of energy, whereas active thermal control requires energy. The active thermal control is simplified for the ÈxypnosSat mission and only an electrical heater is intended.

4.3.5 Telemetry, Tracking and Command System - TT&C

The communication between the spacecraft and the ground station is realized through the telemetry, tracking and command system (TT&C ). The payload data and health status of the spacecraft are transmitted to ground station (also known as telemetry) and commandos from the ground station are received by the spacecraft through the transceiver component. The signal can either be transmitted/received by a high gain (HG) antenna or low gain antenna (LG). A high gain antenna transmits a signal with a higher amplification, but with smaller beam width. As a consequence the antenna has to be directed with high accuracy towards the ground station. Vice-versa a low gain antenna transmits a signal with a broader beam width, but a lower amplification. Usually a spacecraft owns both antennas, since a high gain antenna is required to transmit large amounts of data and a low gain antenna is necessary, in case of emergency (e.g. high gain antenna failed or can not point to ground station due to ADCS failures). Therefore low gain antennas can be seen as backup antennas and should be distributed equally around the satellite in order to be always able to communicate with the ground station during the contact duration. The ÈxypnosSat consists of transceiver, high gain and low gain antenna, whereby transceiver and low gain antenna are double-redundant (same design) and the high gain antenna is not redundant.

4.3.6 Payload

Payloads are required to accomplish the specified mission and are uniquely developed for each mission. It exists several payload types for different mission purposes. Since ÈxypnosSat is an Earth observation satellite, remote sensing payloads are appropriate. The intended remote sensing payload in the invented mission is a slow scan television (SSTV) - camera for imagining earth’s surface and near-earth space. SSTV is a way to transmit static images, in this case, to the ground station. Thereby the images are transfered through the transceiver as audio signal.

The modern SSTV features allow to transmit monochrome images as well color images with

(38)

4.3.6 Payload 27

high quality.

Another payload on-board of ÈxypnosSat is an autonomous on-board decision-making system - ASAP, which is currently been developing at the University of Würzburg. It detects unexpected events and reschedules the plan in order to investigate it. By means of ASAP even short-lived phenomenas will be not missed by the spacecraft, since in common spacecraft missions the operation schedule is changed delayed only by the ground station and only during contact duration [36]. ASAP consists of an imager and planning system. The task of the imager is to detect not predictable events by detecting the changes of captured images. If an event is detected, the ASAP planning system assists by rescheduling the operational plan of the spacecraft [4]. However in the ÈxypnosSat mission there is only one camera implemented for ASAP and observations. ASAP is one of the essential components of the designed decision support system for the ÈxypnosSat. Its task is to detect unexpected events, as described and forward them to the DSS as an input, which will be described in more detail in Section 7.1.

The last payload set in the ÈxypnosSat mission is the Autonomous Diagnosis System for Satellites - ADIA++. Its task is to recognize failures and anomalies of the spacecraft autonomously on-board and to determine their causes. At the moment ADIA++ is been as well developing at the University of Würzburg [37]. It is another essential payload for the design of the decision support system and delivers additional input to it. Details about the input delivered by ADIA++ will follow in chapter 7.

Design of an Autonomous Decision Support System for High-Level Planning in Nano Satellites Using Logic Programming

Design of an Autonomous Decision

Support System for High-Level Planning in Nano Satellites Using Logic Programming

Saliha Serdar

Design of an Autonomous Decision Support System for High-Level Planning in Nano Satellites Using

Logic Programming

Master Thesis in the course of the study programme

"Master in Space Science and Technology" by

Saliha Serdar

Statutory declaration

Contents

Abstract

Acknowledgment

Acronyms

1 Introduction

2 State of the Art

2.1 On-Board Autonomous Science Investigation System for Opportunistic Rover Science - OASIS

2.2 Autonomous Exploration for Gathering Increased Science - AEGIS

2.3 Autonomous Science Target Identification and Acquisition - ASTIA

2.4 Multi-Rover Integrated Science Understanding System - MISUS

2.5 Autonomous Sciencecraft Experiment - ASE

2.6 Project for On-Board Autonomy - PROBA

2.7 Conclusion of the State of the Art

3 Theory

3.1 Definition of Decision Support System - DSS

Rule-Based Decision Support System - Expert Systems

3.2 Logical Programming Language - Prolog

3.3 Analytic Hierarchy Process - AHP

Intensity of Importance Definition Explanation

3.3.1 Detailed Approach of the Analytical Hierarchy Process

3.3.2 Super Decision Software

3.3.3 Advantages of AHP over the Simple Scoring Model

4 Spacecraft Mission Design

4.1 SONATE

4.2 Orbital Design

4.3 Spacecraft Subsystems

4.3.1 On-Board Computer - OBC

4.3.2 Power System

4.3.3 Attitude Determination and Control System - ADCS

4.3.4 Thermal Control System

4.3.5 Telemetry, Tracking and Command System - TT&C

4.3.6 Payload