WSN Configuration using Agent Modeling and Hybrid Intelligent Decision Support System

Transcription

1 Internatinal Jurnal f Sciences: Basic and Applied Research (IJSBAR) ISSN (Print & Online) WSN Cnfiguratin using Agent Mdeling and Hybrid Intelligent Decisin Supprt System F. Alshahrany *a, H. Zedan b, Idir Mualek c a,b Schl f Cmputer Science and Infrmatics (STRL), De Mntfrt University,The Gateway, Leicester, LE1 9BH,United Kingdm c Intelligent Decisin Systems, Jerjer Huse, Nttingham, United Kingdm a falah88@htmail.cm b hussein.zedan@ggl .cm c imualek@yah.c.uk Abstract A cnceptual multi-agent framewrk based n a knwledge-based cllabrative decisin supprt is prpsed in this paper t design hybrid intelligent decisin supprt systems (HIDSS) based n plicy settings fr the supprt f intelligent and pr-active decisin making activities. An example f HIDSS has been develped t supprt the design and cnfiguratin f small wireless sensr netwrks (WSN). A WSN prttype is designed in this research t supply real time envirnmental and cntext related data and decisin knwledge t emergency respnse applicatins specifically in the dmain f civil defense. Its integratin in the HIDSS is essential fr the interactive supprt f all the decisin making prcess phases, and will cntribute t enhance the quality and the scpe f autmated individual and grup decisin making in emergency preparedness and respnse applicatins. This autmatin is based n the cncept f Just In Time Knwledge Management (JITKM) enabled by the use f WSNs. The cnceptual framewrk is firstly presented after a brief definitin f the prblem and design requirements fcusing n bth rganizatinal and technical perspectives, in the cntext f inter-rganizatinal activities perfrmed in a cmplex and dynamic management envirnment * Crrespnding authr. address: falah88@htmail.cm. 67

2 Then the slutin develped fr the design and cnfiguratin f the WSN is succinctly described. It cnsists f eliciting the sensing requirements translated int the hmgenus sensr nde specificatins, lcalizing these sensr ndes using a pre-planning prcess and cnfiguring them prir t their deplyment. The cnfiguratin characteristics f the hmgenus and hetergeneus sensr ndes are presented in terms f plicy settings fr the definitin f the WSN architecture. Results frm the case study are finally presented t illustrate the implementatin f the slutin in emergency preparedness fr fire detectin. Keywrds: Hybrid intelligent decisin supprt system; Knwledge-based cllabrative decisin supprt; Wireless sensr netwrks; Service cmpsitin and rchestratin; Multi-agent systems; Intelligent agents. 1. Intrductin Emergency preparedness and respnse decisin making are an essential aspect f emergency management. The decisins may invlve difficult steps f ad hc decisin-making tasks perfrmed in islatin r cperatin by individuals r grups f peple. Every individual r grup f peple might have different perceptins f the prblem and prvide different plausible and feasible slutins [1]. The difficulty in perfrming these tasks in ad-hc situatins can be accentuated by the varied nature f the emergency typlgy mainly when its cmplexity is characterized by the extent f the disaster. The disaster cntext has several aspects that include the hazard, the scpe, the site, and the prevailing cnditins. These aspects can be mdeled as cmpsed cmplex hetergeneus agents f the applicatin dmain, interacting differently with each ther, depending n their scpe, their cnditins f influence, and the extent f the area r subject matter they are assciated t, and the cnditins f their invcatin. In the prpsed cnfiguratin setting t mnitr the activity f the applicatin dmain, sensr ndes cnnect wirelessly multiple autnmus smart devices having different infrmatin and/r diverging interests. The hetergeneus agents f the applicatin dmain abve mentined, the smart devices f the setting including the detectrs and the hetergeneus devices detailed later in this wrk, and their sftware intelligent agents required fr their deplyment, cnfiguratin and cntrl cmpse the multi-agent system. The emergency respnse can be ne in a kind, making even harder its planning and executin. This particular specificatin f the emergency respnse slutin requires three essential characteristics f the multi-agent system: lgical knwledge acquisitin and representatin, multi-agent cmmunicatin and ratinal agency [2]. The fast grwing WSN technlgy has impsed new challenges n netwrk design, such as t imprve the cnnectivity between sensr ndes, small adhc WSNs and ptimal message ruting t integrate hetergnus devices and incrprate ther intelligent devices using multi-agent systems. This integratin must enhance the perfrmance f the verall efficient decisin supprt, and the verall system must be flexible enugh t adapt t changes rapidly by integrating individual r cmpsed agents. This step takes the existing netwrks twards the generalizatin f fully autmated true web-based netwrks [3,4], supprted by a variety f different types f supprt systems intelligence and knwledge riented, and agent-based mdel t study the behavir f cmplex systems. These Supprt systems integrating varius intelligent cmputing techniques based n advanced intelligent cmputing theries and multi-agent systems, have been widely develped fr many different 68

3 applicatin dmains t slve intractable prblems invlving cmplex mdeling, ptimizatin and genetic prgramming. These techniques include AI technlgies such as fuzzy lgic, rule inductin, neural netwrks, genetic algrithms, case-base reasning, hybrid artificial intelligent systems, data-mining algrithms, intelligent agents and multi-agent systems, expert systems [5]. Cntemprary intelligent cmputing techniques intrduced the cncept f agent and web-service cmpsitin t enhance the generatin f intelligent agents advcated t bridge the cmputatinal intelligence, knwledge discvery in database, decisin supprt and intelligent cmputing technlgies in the supprt perspective f JITKM [6]. This paper is rganized as fllws: next sectin briefly discusses the prblems faced in the design f a WSN based HIDSS with a particular emphasize n the key issues f the prblem dmain. Related wrks are presented in sectin II, and the cnceptual framewrk is described in Sectin III after the examinatin f the framewrk requirements and the strategy. Sectin IV details a case study based n a multi-rm building. Lastly, the cnclusin f this wrk is presented in sectin V. 2. The prblem Disaster management, which invlves a huge number f hetergeneus agents in a hstile envirnment, is supprted by emergency respnse systems depending upn time critical and detailed infrmatin t make real time decisins. Such systems are: cnfigured as ptimal hybrid intelligent decisin supprt systems t make crrect and quick decisins at all the disaster stages, supprted by WSN s t prvide efficient, lw pwer and fast cmmunicatin mechanism t cllate critical decisin data and transfer it t update the cre knwledge f the HIDSS. The design f such systems requires taking int accunt the fllwing cnstraints: a) The fast cmmunicatin f critical data frm the WSN t the central cntrl system depends upn many but imprtantly n tw factrs: available bandwidth and netwrk traffic. The netwrk traffic lad is minimized using data cntrl and prcessing at the active sensr nde level. The main research questin is t hw much lcal prcessing shall be perfrmed as nt t verburden the limited capabilities f the WSN ndes. b) The synchrnizatin f decisin making data generated frm different wireless netwrk surces in a multiagent system and the adaptatin f its agents is a cmplex decisin task aimed at generating sme meaningful results in the cntext f knwledge discvery. This knwledge is essential t supprt multiple plicy making prcesses t elabrate strategic decisins in the cntext dmain f the use f the WSNs. The decisin making prcess knwledge supprt is a key cmpnent in the design f HIDSS aimed at elabrating the sensr ndes intelligence needed t dynamically adapt their behaviur and apprpriately react t the changes identified in the cntext envirnment. 69

4 c) A federatin f autnmus intelligent agents assures the dynamics f the WSN. These agents use sensing and elabrated data made available frm the netwrk and the crprate database t identify new changes ccurring in bth the netwrk and the envirnment cntrlled by the multi-agent distributed system. d) Agents interactin in multi-agent systems reflecting the changes in the dynamic agent envirnment requires negtiatin mechanisms t supprt the incrpratin f causal relatinships between structured negtiatin terms (SNT) and unstructured negtiatin terms (UNT) in the prcess f agent negtiatin and cnflict reslutin. SNT and UNT are a cncept used in this research framewrk t represent the influence f variables n ther variables, all characterised by an influence directin and magnitude. The mutual impact f intelligent agents characterised by plausible causal relatinships when established between SNT and UNT, induces knwledge and inference rules that can be represented in a cgnitive map [7] using intelligent agents t supprt the autmatic negtiatin mechanism between agents. 3. The Mtivatin HIDSS are advcated in this research t supprt cmplex integrated decisin tasks aimed at adapting decisins t changes in cllabrative decisin making envirnment. These changes are based n the decisin making knwledge gained in real time frm a brad typlgy f the dmain data, thrugh the deplyment f autnmus intelligent agents representing bth intelligent smart devices, humans and sftware agents, all invked individually r in small grups t react accrdingly t the desired targets. The agents interactin are the results f decisin making cmpsed prcesses called cmpsed decisin services which integrate agents f bth cnfiguratin setting and dmain applicatins with intelligent agents, inducing fuzzy relatinships t prvide explanatins f the reasning prcesses. A natural language is therefre needed at the external level t enable the system end-users t define the system agents and elabrate the decisin making mdels t prduce internal r lgical decisin making cmpsed services effectively distributed and simultaneusly accessible t the system agents. The deplyment f multi-agent system prcures autmatic and instantaneus service capabilities fr lgical knwledge acquisitin, t integrate the different decisin making levels, enabling them t integrate, alter and cmpse bth technical and business prcesses supprting the different rganisatinal scenaris. The abve mentined design specificatins are derived frm bth the dmain applicatins requirements and the characteristics f the direct and indirect interactive supprt needs t grup and individual decisin making styles, partly r fully autmated. These characteristics are bradly cmmn t a wide variety f decisin making prblem dmains, and cnstitute the theretical basis frm which the architectural frms supprt mechanisms are cnstructed. The adaptive and flexible design and develpment strategies used t smthly integrate technical and rganisatinal aspects including the cgnitive styles, are supprted by the cnduct f several evaluatin appraches. 4. Related Wrk The fast grwing number f virtual enterprises has a substantial impact n emerging applicatins f the new paradigm f HIDSS. These applicatins resulted frm studies based n using different technlgies t develp 70

5 capabilities that enable rganisatins t adapt t a new rapid changing hetergeneus envirnment characterised by multi-level inter-rganizatinal interdependencies. Related wrk includes several studies and prjects based n different cnceptual framewrks and system architectures integrating new analytical web-based applicatins. These studies have been carried ut specifically in the dmain f HIDSS applied t cmplex systems [8] and in varius dmain applicatins such as medical diagnstic [9], water resurce management [10], business and marketing strategy [11,12], aviatin weather frecasting [13], and natural disaster management [14]. Prjects fcusing mre specifically n WSNs design and cmmunicatin issues include: The Harvard MteLab prject, a web-enabled sensr netwrk Testbed [15], The SLAM prject, describing a netwrk architecture integrating a huge number f sensr actuatrs and distributed sftware applicatins fr rapid disaster respnse, scalable crime detectin and preventin and asset mnitring and navigatin[16], The TriSentinel prject, aimed at designing the first emergency respnder wireless cmmunicatin system (FRCS) that supprts inter-agency and intra-agency cllabratin in emergency respnse between plice, fire, emergency medical services and ther civil defence emergencies [17]. The RbCup Rescue Agents Simulatin prject aimed at prviding a research platfrm fr develping and testing strategies and cmprehensive simulatin systems fr emergency search and rescue in disaster management [18]. 5. Cnceptual framewrk The main fcus in the prpsed cnceptual multi-agent framewrk is the design specificatins f HIDSS characterized by the shift frm prviding the traditinal supprt based n analytical tls and techniques t prduce and evaluate the best decisins ptins, t an intelligent gateway fr the cmprehensive and adaptive supprt f specific knwledge needs [19]. This supprt is essential t build descriptive and predictive mdels needed t perfrm interactively the aggregatin f previus infrmatin and elabrated knwledge t frm the dmain knwledge. The cnsistence f the dmain knwledge, which is cmpsed f factual knwledge and expert knwledge, is a key requirement fr the imprvement f plicy making prcesses generating the strategic decisin fr the WSN cnfiguratin. The dmain knwledge supprts the HIDSS capabilities, mainly the predicting capability which cnsists f a reasning based n a classificatin [20] requiring envirnmental data. The predictin capabilities are built upn the use f a set f situatins r bservatins t interpret the ccurrence f an event within a class spectrum. An example f such capabilities is the decisin refinement perfrmed t reduce the decisin uncertainty between the ccurrence f fire cnditins indicated by sensing data ranging beynd the threshld values r mal-functining sensing netwrk devices in the dmain f fire preventin and detectin. 71

6 Evlutinary develpment in mnitring and data acquisitin in envirnmental dmain applicatins, has impsed the definitin f smart envirnments based n the use f wireless data cllectin and distributin netwrks. Distributed real time systems are using wireless technlgies fr the supprt f varied indr mnitring and cntrl applicatins in buildings, hmes and shipbards. A hybrid WSN supprted by an HIDSS system fr its cnfiguratin and management has been prpsed in this wrk t illustrate the main cncepts examined in this research. The main design requirements fr the design f the system as integrated in the cnceptual multi-agent framewrk are belw presented. a. Requirements The main design requirements are: 1) The tasks distributin in the multi-distributed system envirnment; this distributin needs t clearly separate the capture f envirnmental and cntext aware data and then its cmmunicatin at the netwrk level and then finally turning this int knwledge and decisin data by the supprt system using the inference mechanisms f the HIDSS in the cntext f JITKM. 2) The ptimal use and security f WSNs and the imprvement f its perfrmance ver time and reliability is t be cnsidered upn addressing the majr issues mentined in the research cnceptual mdel, mainly: The distinctive use and imprtance between hetergeneus and hmgeneus devices used in cmbinatin t enhance the decisins refinement in terms f analyzing and interpreting the sensing data utcmes by invking intelligent agents t perfrm inherent decisin making prcesses t identify the best strategy and derive the apprpriate actins t be taken in the cntext f the multi-agent system service cmpsitin. The imprtance f lcal data prcessing at the nde level and its impact n the WSN perfrmance in terms f netwrk cngestin avidance and prcessing delay reductin, and c-peratin with the distributed client/server architecture fr the sensr ndes deplyment, cnfiguratin and management, and messages prcessing. The balance between sensrs and sensr ndes duplicatin within the WSN and their switching ff when apprpriate in the case f multiple sensing, t ensure a high reliability and perfrmance f the netwrk by: duplicating sensible sensrs in strategic sensr ndes, adding additinal sensr ndes r reduce the sensing spacing distance while lcalizing the sensr ndes using active sensr ndes t lcally prcess the sensing data using mbile agents rather than bringing them t the central prcessr [21], and reducing the number f redundant sensr devices and messages. The imprvement f the message ruting mechanism, addressing the key prblems f: 72

7 energy cnsumptin and residual energy in sensr ndes ruting path, and link quality requiring a thrugh instantaneus mnitring f WSN. The aut adaptatin f the WSN t external events frm the envirnment requires mnitring and measurement f criteria inherent t its deplyment, cnfiguratin, recnfiguratin, cmmunicatins imprve its perfrmance. 3) The cmpsitin requirements f services f bth cntext applicatins and WSN t determine the specificatins f cmpsed services t supprt: The netwrk cnfiguratin and cntrl, The facilitatin f the emergency respnse and The planning f emergency actins using intelligent sftware agents and mbile devices. 4) The implementatin issues f system prttypes aimed at integrating several technlgies cntrlled by cmpsed agents and services using a hybrid apprach. b. The Strategy The fcus in this study is nt n the pint-t-pint and multi-drp netwrks, but n the web service netwrk tplgy fr agent deplyment in a distributed cnfiguratin. This architecture assumes that all the intelligent sensr ndes are always wirelessly cnnected, and cperate with the aid f intelligent agents t meet their targets, prduce relevant infrmatin t memrize and cntribute t knwledge acquisitin needed fr ther agents, increase the system autnmy and reduce the cmputing lad n the hst. The agents cperatin results in additinal netwrk capabilities distributed between the hst and the WSN, and fr the WSN between the gateway and the sensr nde t enable self-hsting netwrks including selfcnfiguratin, and effective cmmunicatin and data exchange between agents. The strategy adpted in this cnceptual framewrk is t clearly dissciate the netwrk frm the client functins t enable a functinal separatin f the technical system services frm the business services; with a view f implementing an adaptable and flexible service and agent cmpsitin t integrate the different business prcesses and intelligent agents. c. The multi-agent framewrk The definitin f the cnceptual framewrk is in the cntext f smart envirnments supprted by a distributed multi-agent system. These smart distributed envirnments are cmpsed f three separate levels: rganizatinal, functinal and lgical. These levels are structured and supprted by a distributed multi-agent system where the different entities cmpsing each level are mdeled as agents assciated t ne f the several agent mdeling paradigms [22]: simple reflex, mdel r gal r utility based, learning, decisin, prcessing, 73

8 input and physical. Each agent representing a physical r lgical entity has different degree f perceived intelligence and capability, and bserves thrugh sensrs and acts upn an envirnment using actuatrs. The agent actins are the results f the knwledge mechanisms supprted by an inference engine based n the use f an integrated knwledge and mdel base. The agent cnditin-actin is assigned t a realizatin target which might change depending n the knwledge acquired and prduced in the dmain applicatin. d. The distributed multi-agent system The distributed multi-agent system architecture advcated in this wrk is structured t supprt the rganisatinal entities f the activity dmain and their interactin with each ther. These are rganised in terms f services structured in different levels illustrated in Figure 1. These levels t the specificity listed belw, and their interactin represented in the emergency respnse generic functinal mdel shwn in Figure 1. Figure 1: The distributed multi-agent system services levels. The prpsed system architecture aims at supprting the links f the service layers f the dmain applicatins t several integrated layers f the physical cnfiguratin, the WSN, mbile sensing and cmputing devices and technical services required by the system agent services rchestratin. a) The physical level is characterized by a cnfiguratin f distributed smart devices, which takes up sensry data frm the real wrld lcatins. The integrated sensrs f different functinalities are remtely cntrlled at diverse lcatins t acquire data and turn it int knwledge. These sensrs when switched n, perfrms data acquisitin, while distributin f this data ver the netwrks is mnitred and cntrlled by a management center supprted by an HIDSS, which then frward this data t a smart cntrl rm (explained in next sectin). These netwrks vary in their architecture depending n their cmplexity that is based upn envirnmental, dmain applicatin and situatinal factrs. The main factrs are related t the specificity f the sensing, prcessing and decisin making tasks, and the technlgy used in design the netwrks. 74

9 b) The functinal level is characterized by the interactin f the different autnmus entities f the physical level. The entities can be deplyed alne r gruped. c) The lgical level invlves the real time decisin making prcesses and the service rchestratin which includes the service cmpsitin and deplyment f autnmus intelligent agents. d) The real time decisin making mdels include decisin-making tasks that are based n a smth integratin f nn-intuitive analytical slutins t cunter the different prblem aspects f emergency preparedness and respnse. These analytical slutins are advcated in this cnceptual framewrk t meet the fllwing bjectives: Supprt the centralized situatin mdel f the emergency respnse in all its aspects and phases, Imprve the cmputer supprted cperative wrk in the emergency respnse cntrl rm, Enhance the emergency respnse facilitatin prcess, Enhance the quality f the grup decisin making in team based peratins, and Enhance situatinal awareness fr: imprved emergency respnse tactical and peratinal efficiency, emergency services safety and emergency respnse reliability. e. Organisatinal entities The main entities characterising the rganisatin level f the cnceptual framewrk are presented belw. a) The site under surveillance The Site described belw is used in the Case Study mentined in the next sectin. This site is a hierarchical structure f a number f rms gruped t frm a building with smart features and accessries: Each rm is made up f partitins, separated by walls; these partitins have different cnstructin patterns (wall partitin, dr and windw) and characteristics (single, duble, and glass). The Energy supply use pints fr gas and electricity are indicated where existing.walls and wall partitins are identified, such as t assciate characteristics infrmatin t them, t enable the calculatin f the hazard level. This can be derived frm the cnstructin characteristics f the building, in additin t its cnditins f use. Fr e.g. a fire risk assesment mdule is used t assess the fire hazard and measures t emergency preparedness. b) The WSN A small scale WSN is cnsidered in this wrk, and invlves a structured sensr nde placement by hand using a priri planning. Multi-mdal sensr ndes with sensrs n bard are used in this WSN. Structurally, this WSN 75

10 is rganised as a gruping f hmgenus sensrs ndes and hetergeneus devices wirelessly cnnected t a gateway via ruter ndes. Hmgenus sensr ndes: cntain all the sensrs needed fr sensing the detectin elements fr fire predictin and detectin such as temperature, smke, light, gas, etc... Hetergeneus devices: include the entire auxiliary intelligent and smart devices needed t supprt and enhance the activity f the hmgenus sensr ndes; these devices include cameras, mtin tracking, sprinklers, peple cunters and cntrl devices fr autmatic instantaneus clsure and pening f drs and windws. Ruter nde: is an active sensr nde invlved in ruting netwrk messages between gateway and grup f sensr ndes. These ndes maintain a ruting table and manage lcal address allcatin. Due t the devices differences, their integratin in the WSN requires a high-level f data mdularity and adaptability as described in the Multi-Agent System Architecture [23]. This is illustrated in the prpsed cnceptual framewrk Figure 1 by the several service levels f the distributed multi-agent system. c) A smart cntrl rm Based n the use f a vide wall and ther display devices, the cntrl rm is cnnected t several cmputers with multithreading and GPU prgramming t prvide intelligent cntext dependant user interface needed t: Display the site surveillance situatin at any time and shw the results f the risk assessment prcedures, Cnduct emergency respnse scenaris using situatin mdelling [24] in persn detectin and tracking, als fire predictin and detectin t design plausible emergency evacuatin prcedures, and Supprt the cperative wrk f the emergency respnse team and running f the emergency peratins in their different aspects in parallel and real time. d) A cmputer rm Lcating several cmputers with multithreading prgramming fr the strage f data and deplyment f intelligent agents cmpsed services, this rm is cnnected t the smart cntrl rm, mbile cmputing and device cnfiguratin deplyed n sites during emergency respnse, emergency and rescue services and gvernmental agencies. This rm is assciated with all the intelligent agents cmpsing the technical services deplyed in the distributed multi-agent system when supprting the emergency respnse and the implementatin f the different scenaris. e) Emergency and rescue services are plice, fire and rescue service, emergency medical services. f. The functinal mdel The functinal mdel elabrated frm the cnceptual framewrk and shwn in Figure 2, aims at studying the cntext applicatin sensing requirements, defining the specificatins f WSNs and their deplyment cnditins 76

11 t enable the autmatic selectin and recnfiguratin f the WSN sensr ndes clusters wirelessly cnnecting buildings and emergency services. These requirements are turned int specificatins used t identify the needed physical and sftware agents t supprt the different prcesses f the emergency system as illustrated in Figure 3. The agents hierarchy can be elabrated frm the typlgy f functins resulting frm the analysis f the emergency respnse business prcesses. Sub-agents can be used t perfrm lw level functins crrespnding t frmal simple tasks whereas agents will be devted t cmplex tasks. These are a cmpsitin f lw level functins using a hierarchy f sub-agents as explained in the service cmpsitin mdel. Figure 2: WSN cnfiguratin functinal mdel. g. The lgical mdel A federatin f intelligent autnmus agents deplyed in a multi-agent system t supprt the activity structured in the cntext applicatins cntrls the deplyment f these WSNs t smart buildings fr surveillance activities. In this paper, this activity aims at planning emergency preparedness and respnse. This activity cnsists f supprting the need fr detecting temperature rises and fires, identifying, lcalizing and tracking human presence in public attended clsed places such as buildings and imprve the fire detectin and mitigatin decisin-making prcess. The activity f emergency preparedness and respnse is structured in a sequence f prcesses linked by cnditins t be met t establish the triggering events that define the services rchestratin rules, as described in the HIDSS architecture shwn in Figure 3. h. The service cmpsitin mdel 77

12 The cnditins C1, C2, C3 and C4 as shwn abve in Figure 3, and the internal cnditins present in the different cmpnents f the HIDSS are cmpsed using a hierarchy f intelligent agents and sub-agents t perfrm the system functins and lw level functins. These cnditins are the identificatin base f the events, which dynamically link them t the available services created by the cmpsed intelligent agents deplyed in the HIDSS and cnnected thrugh the WSNs. Fr example, the cnditin C1 is based n the identificatin f factual links between the activities f emergency preparedness and surveillance establishing elabrated infrmatin f a nature t prduce knwledge needed t frmulate the knwledge rules cntained in the varius mdels cmpsing the mdel base. This cnditin requires the use data mining intelligent agents during the service cmpsitin prcess. The available services fr a given WSN r applicatin dmain are cntained in a repsitry, while the cmpsitin rules are stred in the IDSS mdel base, whereas the whle cmpsitin scenari fr this service is described in the HIDSS knwledge base. Figure 3: Hybrid intelligent decisin supprt system architecture. 6. The Case study Implementatin a. The case study The Case study presented in this sectin will illustrate the emergency preparedness and respnse system design as prpsed in this wrk. A single building structure under a small WSN surveillance is mdelled using the prcess shwn in Figure 4. In this prcess, the details f the building layut and cnstructin are used t derive cncurrently the evacuatin and surveillance devices requirements. The evacuatin requirements are the basis fr the study f the evacuatin scenaris, whereas the surveillance devices requirements are needed t design the apprpriate sensr ndes by selecting available sensrs and their specificatins. These ndes, nce autmatically lcalised, will be cnfigured, clustered and linked t cmpse the WSN hmgeneus ndes. The WSN similarly cnnects the 78

13 hetergeneus devices which are als autmatically allcated. The WSN hmgeneus and hetergeneus sensr ndes shwn in Figure 6 are cntrlled by intelligent cnfiguratin agents. b. Implementatin 1. Building mdelling The creatin f the building layut is a requirement fr the lcalizatin f hetergeneus devices and hmgenus sensr ndes. The layut is generated frm a building wall segments drawing as shwn in Figure 5, drawn using the directin encding system [25]. Figure 4: Emergency preparedness prcess. Figure 5: Wall segments drawing & Encding system. 79

14 2. Ndes lcalizatin a) Ndes lcalisatin mdel Rules f thumb are used t lcalize the hetergeneus devices (centre f the ceiling and prximity t drs and windws), whereas the central place thery algrithm is used t lcalize the hmgeneus sensr ndes. The ndes lcalizatin mdel is shwn in Figure 6. Figure 6: Ndes lcalisatin mdel. In this case study, tw mdes f nde lcalizatins Hexagnal (Figures 7A) and square gemetric (Figures 7B), have been implemented using different nde lcatin distributin patterns (NLDP). Figure 7A1 Figure 7A2 Figure 7B1 Figure 7B2 Figure 7: Gemetric lcatin distributin pattern. b) Sensr ndes and devices lcalizatin example Using a sensr sensing distance f 3m, the fllwing sensr ndes and devices lcatin is autmatically generated as shwn in Figure 8a and 8b. 80

15 Figure 8a: Hexagnal gemetric lcatin distributin pattern. The gemetrical lcatin distributin pattern is cnsidered in this framewrk as the basis fr the sensr ndes lcalizatin. The cnsideratin f bth patterns hexagnal and square with different settings (lcalizatin per rm r building, spacing distances, different clusters, ) prcures an extensive supprt fr a multi-criteria sensr ndes lcalizatin decisin making apprach. Figure 8b: Square gemetric lcatin distributin pattern. The increase f the number f sensr ndes when the spacing distance is reduced pses the prblem f hw ptimal can the gruping f sensrs in the sensr nde be rganized. Table 1 shws fr example, fr the building layut shwn in Figure 4, thse 7 extra ndes (23.33%) will be required, when 3 m spacing is used instead f 4 m fr the Hexagn pattern. It is thus f interest t cnsider the segmentatin f the spacing distance range t enable the gruping f the sensrs t characterize the sensr ndes. Each segment f this range will crrespnd t a layer f sensr ndes as illustrated in Figure 8 where 3 and 6 m are just examples f segments 81

16 threshld, which can be elabrated when cnsidering the variety f sensrs required fr use in the dmain f the cntext applicatins. Table 1: Relatin between spacing distance and number f sensr ndes. NLDP Hexagn Square Sp dist (m) Number f Sensr ndes Difference Rati Figure 8: Senr nde and sensr status setting. c) Sensing verlapping The cncept f single and duble sensing zne fr hmgenus sensr ndes is used t evaluate the sensing verlapping shwn shaded in Figures 7 and in Table 2. The duble sensing zne (Dsz) is calculated using the fllwing frmula: Hexagn NLDP: Dsz = (1) Square NLDP: Dsz = (2) The single sensing zne (Ssz) is calculated using the fllwing frmula: Hexagn NLDP: Ssz = (3) Square NLDP: Ssz = (4) ( r indicates the sensr sensing distance) 82

17 SN/B sensr ndes allcated ignring the existence f wall between the building rms; SN/R takes int accunt the walls. Table 2: Duble and single cverage ratis. NLDP Hexagn Square Sp Dist S N/R S N/B Sensing Area Duble Sensing Single sensing Rati Duble/S Rati SSz Rati Ssz Relevant infrmatin can be extracted frm Table 2: the rati (Duble/Single) sensing zne is 2.65 times mre imprtant when using the Square gemetric and nly 0.52 time mre imprtant when using the Hexagn gemetric sensr nde allcatin pattern. This suggests the preference fr the square mdel, which results in sensing verlap simultaneusly between tw perating sensr ndes. This sensing verlapping culd result in sme activity redundancy (example f alarm stripping) which if detected can be prcessed at the nde level shuld ne at least f the sensr ndes invlved is active. 83

18 3. Sensr ndes deplyment The deplyment f sensr ndes and their cnfiguratin is a cntinuus peratin in the time requiring the real time mnitring f their wrking cnditins depending n their status and clustering. a) Sensr ndes status The status f sensr ndes is specified depending n its sensing requirements and safety factr. The fllwing status specificatins are as fllws: All time active sensrs ndes are thse lcalised, either at the centre f a cluster f sensr ndes r nearest t the energy supply use pints, drs and windws, whereas duty cycled manner sensrs ndes cncern the remaining nes. Active sensrs ndes are thse lcalised at the centre f a cluster f sensr ndes with additinal functinality t perfrm the netwrk functins at the sensr nde and als data prcessing and strage. Active versus Inactive sensr: the sensr cnfiguratin status f the sensr ndes as described by the prcess mdel in Figure 10. Figure 10: Senr nde and sensr status setting. b) Sensr ndes clustering Sensr ndes clusters (SNC) are required t cnnect sensr ndes t ruter ndes and t distribute the netwrk functins fr sensr deplyment, cnfiguratin, activatin and data prcessing. A SNC is a grup f sensr ndes surrunding a central sensr nde. The prcess f definitin f these SNCs is based n searching all 84

19 plausible clusters, ranking them n the descending rder f their ppulatin, and brwsing the building layut frm the left t the right, and frm the tp t the bttm t select the SNCs ideal candidates using the Cartesian crdinates f the central sensr nde lcatin f the cluster. All the sensr ndes frming the selected SNC are remved frm the remaining SNCs waiting t be selected. A prcess f cluster aggregatin is prpsed t add the selected less ppulated SNCs t the mst ppulated SNCs f their prximity. The sensr ndes clustering is a dynamic prcess which is required when deplying and (re)cnfiguring partly r entirely the WSN. The recnfiguring cnditins are autmatically determined by the cnfiguring agents. A SNC is required t have a sensr nde cluster head (snch) that must be supplemented by a sensr nde cluster head substitute (snchs) in case f the snch malfunctining. The structural unifrmity f the netwrk can be extended t integrate: The netwrk needs t preserve energy by assigning, taking int accunt the sensing requirements at each sensr nde lcatin, the sleeping mde t sensr ndes enabling them t switch between active and sleep mdes depending n the netwrk activity t cnserve energy [26]. The gulf existing between sensing spacing distances (SSD) f sensrs that are needed at the same sensing lcatin, which might result in gruping sensrs f the same SSD in different types f sensr ndes. The prevailing cnditins while using the netwrk and the prbable detected events require migratin, which might suggest the netwrk recnfiguratin. These three structural characteristics can be mdeled arund the cncept f virtual cluster defined at a lgical level whereas the initial clustering made f the aggregated SNCs crrespnds t the physical level, which is assciated t ne, r several lgical levels, as illustrated in Figure 11. Examples f cluster lgical levels are: All the sensr ndes invlved in an emergency respnse cnfigured as ne sensr nde cluster with mre reliable cnnectin specificatins (active cnnectin, high perfrmance ruters, high band signal) whereas the ther sensr ndes f the WSN will be cnfigured separately in ther clusters. All the nn switchable sensr ndes are cnfigured as ne sensr nde cluster whereas the ther sensr ndes f the WSN will be cnfigured separately in ther clusters. 4. Evacuatin planning Evacuatin planning is an essential functin f emergency preparedness. The presence f peple attending a building requires cntinuus mnitring and tracking. This will enable the HIDSS t elabrate the key data needed t check the safe use f the building in terms f real ccupancy, and the evacuatin feasibility given the existence f emergency exits and their dimensins. An example f evacuatin decisin mdel displaying the real rm ccupancy and standard evacuatin times is shwn in Figure 12. An example f evacuatin intelligent agent cnsists f tracking peple identified by RFID tags in a rm by using RFID readers integrated in IP cameras, assisting these peple with vice messages t fllw the evacuatin instructins and path. 85

20 Figure 11: Autmatically generated sensr nde clusters. Figure 12: Example f evacuatin decisin mdel. 5. Messages management system supprt The message management system fcuses n tw key particular issues: message ruting and message authenticatin. Their imprtance depends greatly n the cmplexity f the nature f the nature and the type f WSNs. Intelligent agents cntrl the messaging cnditins fr effective message ruting and authenticatin. 86

21 a) Message ruting Message ruting is a very critical which begins with ndes lcalisatin and neighburhd discvery fr the building up f the lcal neighbur tables, starting first by identifying the sensr ndes and their lcatins, their remaining energy, the quality f their link, and their ruting delays. Then, messages are ruted using a gegraphic frwarding algrithms that takes all int accunt the sensr ndes lcatins f the ruting path. In this framewrk, the small adhc WSN sensr ndes are lcalised using the planned methd f the WSN deplyment, and the main issues addressed in the design f the HIDSS are the message ruting reliability, integratin with wake/sleep schedule, mbility, real-time, security, vids and cngestin [27]. b) Message authenticatin Message authenticatin is a key mandatry service in a message management system fr a secure WSN t reduce message falsificatin and mdificatin as well as nde impersnatin [28]. The WSN security is essential, and the preventin f malicius agents injecting messages t the netwrk withut being detected must be supprted by the WSN management f the HIDSS using cryptgraphic techniques. The WSN messages which are prperly authenticated, are the result f the netwrk activity cnsisting f: The sensr ndes deplyment and nde defeated replacement, The sensr nde cnfiguratin and re-cnfiguratin, The cllectin f the Sensing data, and The netwrk management, setting netwrk and security parameters. The message management system is supprted by a message service cmpsitin f different intelligent agents, as illustrated in Figure The HIDSS cmputatinal mdel The cmputatinal mdel supprting the HIDSS is a multi-level agent-based mdel required t dynamically simulate the actins and interactins f bth hmgeneus and hetergeneus device during the surveillance activity in the smart rm and the prgressin f the emergency respnse and recvery actins. This simulatin aims t assess the deplyment cnditins f the HIDSS autnmus agents and the impact f their actins n ther agents and the emergency respnse system. This simulatin which is an nging prcess, is required fr the real time agents mapping t alter their cnfiguratin, hierarchy and target/gal. This agent mapping is integrated in the HIDSS agent-based mdel. The message management system can be specific t the WSN r integrated in the HIDSS. 7. Cnclusins The wrk presented in this paper is a cntributin t the enhancement f the physical wrld r real-wrld interactin using distributed netwrks f integrated wireless and digital embedded sensr-based and cntrl devices technlgy supprted by an HIDSS. 87

22 Figure 13: Message service structure. This cntributin is als in the dmain f integrating different cmputing techniques t supprt the service prvisin and cmpsitin that requires the develpment f intelligent agents including bth intelligent devices and sftware agents. These intelligent agents are the vectrs f the knwledge discvery and cmputatinal intelligence required by adaptive and flexible decisin mdels fr the prcessing f cntext-aware data using the mutual benefits f the varied service access devices. The majr benefit f this framewrk is the cntrl f distributed small netwrks f activity surveillance in a fully autmated setting required fr the ptimal use f emergency respnse resurces. The use f bth physical and lgical clustering levels prcures a flexible and adaptable plicy making supprt t varius rganizatinal scenaris fr different types f emergency respnses. Acknwledgment We wuld like t thank the Saudi High Educatin Ministry and the University f DeMnfrt (U.K) fr the supprt f the research grant, and Jerjer HIDS fr their sftware develpment cntributin. References [1] Viviane B. Diniz, et al, (2008) Decisin Making Supprt in Emergency Respnse Encyclpedia f Decisin Making and Decisin Supprt Technlgies, pp [2] Leytn-Brwn K., and Shham Y., (2009) Multiagent Systems: Algrithmic, Game-Theretic, and Lgical Fundatins. Cambridge University Press,

23 [3] Ammari,H., and Sajal K. Das, (2008) Prmting Hetergeneity, Mbility, and Energy-Aware Vrni Diagram in Wireless Sensr Netwrks July 2008, In IEEE Transactins n Parallel Systems Vl. 19 N. 7 pp [4] Krishnamachari B., (2005), in Netwrking Wireless Sensr: Deplyment & cnfiguratin Lcalizatin Synchrnizatin Wireless characteristics, pp 71-72, Cambridge University Press, isbn [5] Viademnte S (2004) A Hybrid Mdel fr Intelligent Decisin Supprt: Cmbining Data Mining And Artificial Neural Netwrks. Faculty f Infrmatin Technlgy, Mnash University, Australia. PhD. Thesis. [6] Nabie Y. Cnteh and Guisseppi A. Frginne, (2006) A Simulatin Study f Just-in-time Knwledge Management (JITKM). In Intelligent Decisin-making Supprt Systems: Fundatins, Applicatins and Challenges, ISBN-10: pp [7] Tlman, E. C. (1938). The determinants f behavir at a chice pint. Psychlgical Review, 45, pp [8] Filip, F.G. (2008). Decisin Supprt and Cntrl fr Large-Scale Cmplex Systems, Annual Reviews in Cntrl (Elsevier) 32 (1), ISSN: [9] Burstein, F. Et al, (2006) A Rle fr Infrmatin Prtals as Intelligent Decisin Supprt Systems: Breast Cancer Knwledge Online Experience. In Intelligent Decisin-making Supprt Systems: Fundatins, Applicatins and Challenges, ISBN-10: , pp [10] H. Kazeli, H, et al, (2002) An Intelligent Hybrid Decisin Supprt System fr the Management f Water Resurces. In the FLAIRS-02 Prceedings. pp [11] Shuliang Li (2000) The develpment f a hybrid intelligent system fr develping marketing strategy. In Decisin Supprt Systems, Vlume 27, Issue 4, January 2000, pp [12] Martinez Lpez, F. and Cassilas, J., (2009) Marketing Intelligent Systems fr cnsumer behaviur mdelling by a descriptive inductin apprach based n Genetic Fuzzy Systems. In Industrial Marketing Management, Vlume 38, Issue 7,Octber 2009, pp [13] Viademnte S, and Burstein F (2001) An intelligent decisin supprt mdel fr aviatin weather frecasting. Advances in intelligent data analysis: 4th internatinal cnference (IDA 2001), pp Cascais, Prtugal. Berlin, Germany: Springer-Verlag. [14] Cica, M et al. (2009). Infrastructure and System Prgramming fr Digital EcSystems used in Natural Disaster Management, 3rd IEEE InternatinalCnference n Digital Ecsystems and Technlgies, Istanbul, TURKEY [15] Harvard Sensr Netwrk Testbed 89

24 [16] Balakrishnan, H., et al, (2001) ITR: Scalable Lcatin-Aware Mnitring (SLAM) Systems, Labratry fr Cmputer Science & Department f EECS, MIT Nv [17] Burkley et Al, (2006) FRCS US Patent US B2, 25 th April [18] Kitan, H. and Tadkr, S. (2001). RbCup Rescue: A Grand Challenge fr Multi-agent and Intelligent Systems. AI Magazine, 22(1) pp [19] Viademnte, S., and Frada Burstein, F. (2006) Frm Knwledge Discvery t Cmputatinal Intelligence: A Framewrk fr Intelligent Decisin Supprt Systems. In Intelligent Decisin-making Supprt Systems: Fundatins, Applicatins and Challenges, ISBN-10: , 2006, pp [20] Viademnte S (2004) A Hybrid Mdel fr Intelligent Decisin Supprt: Cmbining Data Mining And Artificial Neural Netwrks. Faculty f Infrmatin Technlgy, Mnash University, Australia. PhD. Thesis. [21] M. Chen, et al (2006). Mbile Agent Based Wireless Sensr Netwrks, Jurnal f Cmputers, VOL. 1, NO. 1 pp 14-21, APRIL [22] S. J. Russel and P. Nrvig (2003).Artificial Intelligence: A Mdern Apprach, Upper Saddle River, New Jersey: Prentice Hall, [23] Hill, J., Szewczyk, R., W, A., Hllar, S., Culler, D., Pister, K. System architecture directins fr netwrked sensrs. ACM Sigplan Ntices 35(11), pp [24] M. Martinez, R. Stiefelhagen (2013) Autmated Multi-Camera System fr Lng Term Behaviral Mnitring in Intensive Care Units, In Prc. f IAPR Cnference n Machine Visin Applicatins, Kyt, Japan, May [25] I. Mualek,(1997) Intelligent Decisin Supprt Systems fr Interactive Decisin Making in Cmplex Envirnments Applied t Reginal Planning. PhD Thesis, University f Nttingham, [26] Yunis, O., (2006), "Nde Clustering in Wireless Sensr Netwrks: Recent Develpments and Deplyment Challenges", IEEE Netwrk, Vlume 20, N. 3, pp [27] Stankvic, J., (2006) Wireless Sensr Netwrks University f Virginia, Charlttesville, Virginia [28] Rsales, et al, (2009) Efficient Message Authenticatin Prtcl fr WSNIn WSEAS TRANSACTIONS n COMPUTERS Issue 6, Vlume 8, June 2009, pp