Adaptive Decision Support System in Network Centric

DOI: 10.15199/13.2018.7.10

Adaptive Decision Support System in Network Centric Warfare Process (Adaptacyjny system wspomagania decyzji w procesie walki sieciocentrycznej) dr hab. inż Janusz DUDCZYK1, mgr inż. Łukasz RYBAK2 WB Electronics S.A., ul. Poznańska 129/133, 05-850 Ożarów Mazowiecki, Poland Państwowa Wyższa Szkoła Zawodowa w Skierniewicach, Instytut Informatyki i Matematyki Stosowanej ul. Batorego 64C, 96-100 Skierniewice, Poland 1 

2 

Abstract The article presents a concept of Adaptive Decision Support System (ADSS) which makes it possible to realize a network centric warfare strategy. The introduction of this article presents the significance of data processing by information and communication systems on the computerized battlefield. Also, it consists of a description of the process which aims at gaining knowledge from relational databases in terms of building situational awareness and organizing informational advantage. The strategic thought cycle in terms of building an adaptive information and communication system is presented in this work as well. A further part of the article describes the general architecture of decision support system applications, points out and defines their key elements. Finally, an original concept of decision-making model in adaptive DSS is suggested as an idea in which k-Nearest Neighbours classificator and the method of machine learning are used to realize the assumption concerning the adaptation of the system to dynamic conditions on the battlefield. The direction of research as well as possibilities of potential solution optimization are also indicated in this article. Keywords: Electronic Warfare, Decision Support System, Network Centric Warfare, k-Nearest Neighbours Algorithm, Instance-based Learning Streszczenie W artykule zaproponowano koncepcję adaptacyjnego systemu wspomagania decyzji ADSS (ang. Adaptive Decision Support System) pozwalającego na realizację strategii walki sieciocentrycznej. Na wstępie przedstawiono istotę przetwarzania danych przez systemy teleinformatyczne na zinformatyzowanym polu walki. Scharakteryzowano proces wydobywania wiedzy z relacyjnych baz danych w kontekście budowania świadomości sytuacyjnej oraz organizowania przewagi informacyjnej. Przedstawiono cykl myślenia strategicznego w kontekście budowania adaptacyjnego systemu teleinformatycznego. Następnie opisano ogólną architekturę aplikacji wspomagających decyzję, wskazano oraz scharakteryzowano kluczowe ich elementy. Ostatecznie zaproponowano autorską koncepcję modelu decyzyjnego w  adaptacyjnym DSS wykorzystującą klasyfikator k-najbliższych sąsiadów oraz metodę uczenia maszynowego, co pozwoliło na realizację założenia dotyczącego przystosowywania się systemu do dynamicznych warunków pola walki. Ponadto wskazano kierunek badań oraz możliwości ewentualnej optymalizacji rozwiązania. Słowa kluczowe: walka elektroniczna, system wspomagani decyzji, walka sieciocentryczna, algorytm k-najbliższych sąsiadów, uczenie z przykładów

The term ‘net’ in the Network Centric Warfare (NCW) should be comprehended as creating connections (relations) between all subjects taking part in the operation in order to share information and provide each other direct cooperation. It is impossible to gain a  network centric ability without different use of innovative technical solutions i.e. online techniques (including information safety mechanisms), meeting the needs of battlefield mobile users (by introducing programmable and also intelligent structures of Software Defined Radio or Cognitive Radio) and introducing as well as improving different types of wireless protocols for self-organizing and decentralized Mobile Ad-hoc NETworks (MANET) [1]. The key elements which make it possible to receive a network centric battlefield, thus situational awareness in network centric works are creating safe and effective mechanisms of data acquisition, carrying out the analysis, synthesis as well as aggregation, sharing information on every higher level of data processing, creating Knowledge Bases (KB) and using them effectively in Electronic Warfare (EW). [2, 3]. The Decision Support System (DSS) in network centric activities is based on knowledge and describes chosen methods and techniques of data/information/knowledge management and how significantly their use in the process of commanding maximizes the effectiveness of such a system. Elektronika 7/2018

The base for the solution, presented in this article is the use of an example of nonparametric regression k-Nearest Neighbours algorithm in the classification process, on the basis of a few chosen distance measures and feature vector similarity, which are described in the conclusion module.

Data processing in an adaptive decision support system In a  computerized world it is possible to record every single event or state in a  form of ordered data structures. In order to dominate on the network centric battlefield it is necessary to process them efficiently. The article presents a  concept of Adaptive Decision Support System (ADSS), which realizes ‘the data mining’ process. The superior idea here is to use the computing power of IT devices to search for patterns in the data warehouses, which would not be possible to realize by a human because of time and memory limit. ‘Data mining’ is one of the stages of Knowledge Discovery in Databases (KDD) [4]. The use of KDD supports the process of receiving common awareness on the battlefield by defining the current possibilities and existing limits. Furthermore, knowledge discovery with the use of Adaptive Decision Support System improves the speed of commands and the pace of operational activ-

39

ities. This is the result of the fact that in the presented concept, each soldier on the battlefield has direct access to ADSS thus, in a short time he is given an overview of the situation and a  quasi-optimal conduct algorithm. An example of such a solution may be Modular Integrator for C4I (Command, Control, Communications, Computers and Intelligence) Soldier’s System, a device characterized by functional polymorphism, which improves the soldier’s decision making process on the battlefield and increases his situational awareness [5]. The constant access to fresh information makes it possible to create an image of the battlefield, which supports the task adaptation process to ever-changing conditions of the environment, as a  result maximizing the soldier’s situational awareness. The operations described above lead to an increase in informational advantage [6]. The operation of data processing in the information and communication systems may be divided into four stages between which there is a close correlation. The lowest stage in the continuum consists of data as data is a formal description of particular observations and extent of an object or an incident. The data mentioned above consists of ever-changing ‘information’- which can be of different values from a particular bracket of real numbers placed on a number line and discrete ‘information’ which may be features of objects concerning quality, the type of prevalence or information about the lack of some features or elements. Such data may have discrete values, which generally have no specific range of changes on a number line. These describe non-derivative elements of the object structure with the use of a word, a sentence or a sequence of codewords. Numbers are usually used to present data however, they have no particular meaning for the receiver and are not enough for a communication process to appear. These numbers are of different values from a  particular bracket of real numbers and in the geometric interpretation are usually shown as a point in the space of features. In the presented solution, the extracted data is presented in the form of vectors which in a  static way describe the operational state. These gain meaning as a result of normalization, processing, grouping and visualization. The processes of transformations above provide information, which are the interpretation products of ordered data structures. Such information are formulated in the form of fundamental conclusions thus, sharing them is more transparent in comparison with ‘raw data’. The information combined with experience, awareness and familiarity with certain facts create specific knowledge, which encompasses in general, the things the receiver has learnt, deduced and noticed. In the communicative process discussed here, broadening knowledge is the main aim. The highest level in the processing continuum is wisdom. It is a stage when the receiver’s knowledge is so advanced that becomes an expert in assessment. In comparison with knowledge, there is no other direct method for sharing or learning wisdom [4]. In terms of an information and communication system on the network centric battlefield, the operation of data processing needs to be correlated with a General Observe Orient Decide Act (OODA), in which observation relies on collecting data concerning the area, where actions are being taken. The next step is orientation referring to a combination of the collected information with the notional model which determines the decision, which is an aware or instinctive choice of a particular scenario. The last stage in the OODA cycle is an action, within which particular actions are taken according to the algorithm chosen in the previous stage. By definition, the OODA model consists of cyclic operations thus, after the last stage, which is the action, the received new data is transferred to observation stage in the next iteration of this cycle. Such an action deter-

40

Fig. 1. The diagram with the structure of Adaptive Decision Support System. Source: own elaboration Rys. 1. Diagram struktury Adaptacyjnego Sytemu Wspomagania Decyzji. Źródło: opracowanie własne

mines the greatest advantage of the OODA model, its adaptability. In this aspect, it is worth paying close attention to the feedback as what influences directly further data processing cycles is the quality of the environment’s reaction [4, 6].

Architecture of the Adaptive Decision Support System In the literature, there are several definitions describing DSS and diagrams which present the structure of such mechanisms. The architecture of decision support systems is flexible and data flow in such systems results from their original use. However, an essential condition to classify a system of applications as a DSS is for a system to realize the most important functionality i.e. to suggest the optimal decision for the defined input data. In order to realize this function, the whole mechanism needs to be equipped with modules responsible for storage and authorized edition of the knowledge base, making a proper (in terms of input data) decision and a clear presentation of the data processing results. Thus, Adaptive Decision Support System in network centric work can be called a relational knowledge repository, on the basis of which it is possible to generate statistically the processed data. Such data help an independent unit on the battlefield carry out optimal ad-hoc action in particular conditions and also receive the evaluation of the earlier made decision. Figure 1 presents a  diagram with the architecture of the above ADSS system in a  network centric working area. It should be noticed that the suggested ADSS concept needs to be introduced in a hermetic environment, on each level of command. Therefore, a central analysis repository should be defined on the basis of real battlefield scenarios. In terms of adaptivity, it needs to be emphasized that the receivers of the suggested solution become simultaneously a source of information. This attitude is determined by the fact that over the next years, different decisions concerning statistically similar situations may be made. Generally, it is the result of the technological advancement. As it is shown in the Figure above, ADSS consists of a relational knowledge base, a decision model and a graphic user interface. Inside the central data repository there should be a division into vectors representing decision rules i.e. generally described experts’ opinions, and occurrence vectors stored according to the assumed knowledge representation method Elektronika 7/2018

which is assumed. The analysis of the decision rules indicates correlation between the input data and the data verified by specialists – these are information provided by a classic expert system. Whereas, the use of occurrence vectors makes it possible to search for cases whose features are common with the currently analyzed situation on the battlefield. The suggested attitude makes it possible to adjust decisions to the current conditions on the basis of earlier actions and their effects. In the described DDS, functional polymorphism of an independent unit on the battlefield needs to be emphasized as the receiver of a  message becomes simultaneously a  source of data. Such a phenomenon is determined by the fact that after receiving an optimal reply from the system, the user assesses its effectiveness and as a  result, he becomes an expert because his opinion has a  direct influence on further cycles of analyses. Thus, there is a certain regularity concerning the condition of the introduced solution’s performance – adaptive decision support system on the network centric battlefield works effectively only if its users make a  constant communication process by Graphical User Interface (GUI). The GUI in ADSS must be clear, not complex and its priority function is fast data input and visualization [7]. The main aim to maintain ADSS well should be constant improvement of central knowledge repository on the basis of opinions and users experience.

Nonparametric regression algorithm in the process of taking optimal actions  Adaptive Decision Support System

The key element in the described system is the conclusion model, decision support in the suggested adaptive DSS concept is based on the statistical analysis of a  particular state of the real world presented according to the assumed representation method. Figure 2 presents a  diagram of decision making process. The data which is input to the decision model by the user is a statistic description of the battlefield state. Then, as part of initial processing, the data normalization and aggregation to an ordered structure are carried out. A vector with features describing the current state appears in the output of the model according to the assumed knowledge representation model. The already formed data structure is transferred to the classificator which uses the nonparametric regression k-NN algorithm. This is a simple algorithm, in which the classification is carried out on the basis of the analysis of the tested sample with each sample of the training set. While designing a classificator, high measurement complexity of the k-NN algorithm is taken into account as it results from the rule of operation [8, 9]. The structure of the system is flexible thus, the

further the development of the knowledge base, the more it is possible to fragment the searching process of the set using several devices. The concurrent analysis of the central repository of knowledge is a method of optimizing the time of the classification process in adaptive DSS. In the suggested system, similarity of samples is defined on the basis of vectors distances in two metric spaces i.e. Euclidean and Hamming (Manhattan) spaces, according to the formulas (1) and (2), where x and y are row vectors in m-dimensional space V. d

d E2 (x,y ) =(x − y ) (x − y)T,



d

d H ( x, y ) =

1 m ∑ x[i] − y[i] , m i =1

∀x,y ∈ V

(1)

∀x, y ∈ V m

(2)

Furthermore, to set correlation between feature vectors x and y, cosine similarity measure is used, according to the formula (3), where x,y ∈ V, while T is a transposition sign of the feature vectors above, [10, 11].

d

hcos (x, y ) =

xyT xxT yy T



(3)

Achieving the assumed goal i.e. to create an adaptive DSS concept is possible by designing the most possibly optimal relational knowledge base structure. In this concept of a system, a central data repository stores vectors representing decision rules (namely statistically described experts opinions) and also occurrence vectors with activities carried out earlier. As a  result, it is possible to acquire an adaptive reaction of the system in the aspect of dynamics in the changes on the battlefield, in the longest time period possible. Adaptivity of the suggested decision support system is also achieved because of the use of artificial intelligence algorithms. The system which is described, uses the machine learning process in order to develop the central knowledge repository. Instance-based learning method is used in this case as it dispenses with the canon of other machine learning methods, which formulate the general description of target function already at the moment of the learning data input. Learning from the examples depends on remembering a particular state, and its functional generalization is not realized until in the classification stage. However, as it is described in chapter 3., the information receiver in the system is simultaneously a source of information thus, the final assessment of a  particular state is verified by a unit, which takes a specific actions on the basis of the received signs. Only then classified data, suggested by the system, is ready to the training set in the further analysis cycle.

Conclusion

Fig. 2. A diagram with the structure of decision model. Source: own elaboration Rys. 2. Diagram struktury modelu decyzyjnego. Źródło: opracowanie własne

Elektronika 7/2018

The presented concept of adaptive decision support system in a  longer-term perspective makes it possible to create a rich knowledge repository and as a  result increase awareness on the battlefield and accelerate actions, which in the end maxime the informational advantage in a network centric battlefield. The essence of the presented idea is to provide system’s adaptivity in a wide spectrum of time. The assumption was for a  system to ‘learn’ changes appearing in the environment. The problem of this process is sudden reorganizations, depending on the size of the data volume, over several cycles – the system may

41

suggest outdated solutions. It results from the functioning of k-NN classificator, which takes into consideration the most resembling situations and from the newly-formed, sorted set chooses these which appear in it most often. A solution to this problem is remembering, in the central data repository, the date of taking an action, further selection and transfer to the classification process only the data which meets a particular condition. The presented attitude also makes it possible to optimize the search of the main knowledge base as the distances between vectors are measured from the data set with less power. The effectiveness of the suggested solution is also determined by the quality of the data collected in the central repository, and it depends on the transparency of the communication process in which the channel is a  user graphic interface.

Referente   [1] Amanowicz M.: (2010) Zaawansowane metody i techniki tworzenia świadomości sytuacyjnej w  działaniach sieciocentrycznych, Wydawnictwo PTM, Warszawa.   [2] Dudczyk, J., Kawalec, A.: (2015) Fast-decision identification algorithm of emission source pattern in database, Bull. Pol. Ac.: Tech. 63 (2), 385–389.

  [3] Dudczyk, J.: (2018) The Concept of ELINT DataBase based on ERD Modelling, Elektronika (konstrukcje, technologie, zastosowania), nr 2/2018, pp. 34-37. ISSN: 0033-2089. DOI: 10.15199/13.2018.2.8   [4] Rybak, Ł., Dudczyk, J., Jezierski, Z.: (2018) The IT sector as an important element of critical infrastructure, Międzynarodowa Konferencja Naukowa „Infrastruktura krytyczna w systemie bezpieczeństwa państwa i społeczeństwa”, Nysa, 24-25 May.   [5] Dudczyk, J.: (2015) Polimorfizm funkcjonalny Modułowego Integratora, Elektronika (konstrukcje, technologie, zastosowania), nr 11/2015, pp. 109-113. ISSN: 0033-2089. DOI:10.15199/13.2015.11.27   [6] Michalewski, E.: (2010) Analiza systemów sieciocentrycznych, Instytut Badań Systemowych PAN.   [7] Dudczyk, J., Zielińska, M., Wachowiak, F.: (2016) Ergonomic convergence of a Modular Integrator in aspect of soldier’s situational awareness on the battlefield. Journal of Ergonomics 2016, vol. 6 (2) ISSN: 2165-7556. DOI:10.4172/2165-7556.1000155.   [8] Mazurek, M.: (2014) Architecture of clinical decision support system using the Big Data concept, Roczniki Kolegium Analiz Ekonomicznych, , vol. 35, pp. 257–271.   [9] Akhil, M., Deekshatulua, B.L, Chandra, P.: (2013) Classification of Heart Disease Using K –  Nearest Neighbor and Genetic Algorithm, International Conference on Computational Intelligence: Modeling Techniques and Applications (CIMTA) pp. 85–94. [10] Duda, R., Hart, P.E., Stork D.G.: (2001) Pattern Classification. Second Edition. John Wiley & Sons, New York. [11] Theodoridis, S., Koutroumbas, K.: (2009) Pattern Recogniotion, fourth ed. Academic Press, USA.

Nagrody konkursu Lider Bezpieczeństwa Państwa 2018 Uroczysta Gala Wręczenia Nagród, której honorowym patronem był Szef Biura Bezpieczeństwa Narodowego minister Paweł Soloch odbyła się 21 czerwca 2018 r. w Centrum Konferencyjnym Wojska Polskiego w  Warszawie. Konkurs Lider Bezpieczeństwa Państwa był adresowany do firm i instytucji oraz instytutów naukowo-badawczych oraz wyższych uczelni, których oferta znajduje lub może znaleźć zastosowanie w zakresie bezpieczeństwa i obronności. Ocenie podlegały innowacyjne wyroby, usługi i zrealizowane projekty.  Po raz pierwszy zostały nagrodzone firmy oraz ośrodki naukowo-badawcze spełniające kryteria innowacyjności firmy na rzecz bezpieczeństwa i obronności. W tej kategorii zostały przyznane odrębne nagrody i wyróżnienia „Lider Innowacyjności na Rzecz Bezpieczeństwa Państwa”. Celem Konkursu było uhonorowanie polskich firm oraz ich produktów, które mogą stanowić wyposażenie służb mundurowych, a także służących usprawnieniu funkcjonowania ich systemów zarządzania. W tym roku ocena dokonań uczestników konkursu została rozszerzona o  promocję prac realizowanych prowadzonych wspólnie przez przedsiębiorstwa przy udziale placówek naukowo-badawczych.  Instytut Technologii Elektronowej wspólnie z  firmą TELESYSTEM-MESKO sp.  z  o.o. otrzymały diamentową nagrodę „Lider Bezpieczeństwa Państwa – 2018” za „Krzemowe fotodiody bliskiej podczerwieni do zastosowań specjalnych”. Autorami opracowania byli mgr inż. Maciej Węgrzecki z zespołami z Zakładu Technologii Mikrosystemów i Nanostruktur (Z02) oraz Zakładu Mikroelektroniki (Z06). Nagrodzone diody są optymalizowanego detekcji promieniowania podczerwonego o długości fali 1064 nm i są przeznaczone do stosowania w systemach naprowadzania laserowego w technice laserowej i amunicyjnej. Rodzina bezzałogowych statków powietrz-nych ATRAX, do zastosowania m.in. w akcjach poszukiwawczych, ratowniczych, monitorowania granic państwa oraz ochrony osób i  mienia. Maszyna może działać w różnych środowiskach na wysokości do 2000  m w promieniu do 10 km, może zabirać ładunek o masie do 15 kg. Została wyróżniona „Diamentową Nagrodą”. Odbierali ją inżynierowie z Zakładu Samolotów i Śmigłowców Instytutu Technicznego Wojsk Lotniczych, bezpo-średnio zaangażowani w prace konstrukcyjne nad nagrodzonymi dronami, byli nimi mgr inż. Wojciech Lorenc (kierownik pracowni BSP) oraz inż. Marcin Chmiel i inż. Michał Mazur. System ORTO-LBNP, jako jeden z trzech wyróżnionych, został uhonorowany nagrodą w III Konkursie „Lider Innowacji w Dziedzinie Bezpieczeństwa i Obronności 2018”, przyznaną przez Fundację „Promilitaria XXI”, wydawcę portalu-mundurowego.pl oraz Narodowe Centrum Badań i Rozwoju. System powstał w ramach projektu pt. „Opracowanie technologii LBNP (ang. lower body negative pressure)

42

do badań i treningu pilotów Sił Zbrojnych RP w warunkach niedotlenienia niedokrwiennego oraz stresu ortostatycznego”, realizowanego przez Konsorcjum Naukowo-Przemysłowe złożonego z Instytutu Techniki i Aparatury Medycznej ITAM, Wojskowego Instytut Medycyny Lotniczej, Instytutu Biocybernetyki i Inżynierii Biomedycznej im. Macieja Nałęcza, Wyższej Szkoły Oficerskiej Sił Powietrznych oraz ETC-PZL Aerospace Industries sp. z. o.o., finansowanego ze środków NCBiR. Radiostacja COMP@N H09 z  firmy RADMOR jest urządze-niem doręcznym działającym w  systemie SDR (Software Defined Radio), opracowanym z  wykorzystaniem wspólnej platformy sprzęto-wej produktów rodziny COMP@N. Wszystkie radiostacje tej rodziny są programowalne, co oznacza, że zmiany ich funkcji dokonuje się wyłącznie drogą zmiany oprogramowania. Dzięki temu charakteryzują się uniwersalnością, elastycznością łatwością dostosowywania funkcji do wymagań użytkownika. IBIS® – opracowanie Przemysłowego Instytutu Automatyki i Pomiarów jest robotem przeznaczonym do działań pirotechnicznych oraz prowadzenia rozpoznania. Po zainstalowaniu dodatkowych urządzeń może być wykorzystywany między innymi do neutralizacji niebezpiecznych ładunków, rozpoznania chemicznego czy działań ratowniczych. Sześciokołowa platforma mobilna z niezależnym napędem na każde z kół sprawia, że robot z  łatwością porusza się w  trudnym i  zróżnicowanym terenie (podłoże skalne, tereny podmokłe i  grząskie, rumowiska). IBIS® jest robotem szybkim (10 km/h). Konstrukcja zawieszenia bazy mobilnej zapewnia optymalny kontakt kół z podłożem, a co za tym idzie sprawne pokonywanie nierówności terenu, dużą stabilność podczas jazdy oraz właściwe rozłożenie mocy na poszczególne koła. Dzięki tym rozwiązaniom IBIS® charakteryzuje się zwrotnością i  mobilnością. Manipulator z  wysuwnym ramieniem gwarantuje zasięg ponad 3 m i duży zakres ruchu w każdej płaszczyźnie. Za pomocą manipulatora można podejmować i przenosić ładunki o masie do 50 kg. Wojskowy Instytut Techniczny Uzbrojenia został nagrodzony za interesujący produkt: rodzinę głowic bojowych stanowiących uzbrojenie bojowych bezzałogowców. Są to głowica kumulacyjno-odłamkowa GK-2 HEAT, odłamkowo-burząca GO-2 HE i dwustopniowa głowica termobaryczna GTB-2 FAE. Wymienione głowice mogą być instalowane w dronach, jako tzw. amunicja krążąca – kategoria bezzałogowców „kamikadze”; maszyna ulega zniszczeniu w momencie ataku. (cr)

Elektronika 7/2018