36 Journal of Digital Information Management Volume 13 Number 1 February 2015

WSN - Based Indoor Location Identification System (LIS) Applied to VisionRobot Designed By Fuzzy Neural Network

Yi-Jen Mona

a Department of Computer Science and Information EngineeringTaoyuan Innovation Institute of Technology, Chung-LiTaoyuan, 320, Taiwanmonbuy@tiit.edu.tw

Journal of DigitalInformation Management

ABSTRACT: In this paper, the adaptive network basedfuzzy inference system (ANFIS) based wireless sensornetwork (WSN) is constructed to develop a locationidentification system (LIS) to monitor the values of linkquality indicator (LQI) or received signal strength indicator(RSSI) of IEEE 802.15.4/ZigBee WSN indoor to realizethe tag locations of vision robot. The LIS system iscomposed of one gateway, some location nodes and tags.The RSSI methodology is used to identify the location oftags. At first, the LQI is demonstrated then it is appliedto develop a WSN LIS. By using LIS, the performance ofindoor location identification of vision robot is verified. Atlast, the ANFIS neural network is used to combine withthe LIS to develop application of vision robot. Theexperimental results demonstrate that the designrequirement can be achieved; meanwhile, it alsodemonstrates that the good indoor LIS performances arepossessed by using LIS system and ANFIS neuralnetwork.

Categories and Subject DescriptorsC.2.1 [Network Architecture and Design]; Wirelesscommunication: I.2 [Artificial Intelligence]: Uncertainty,``fuzzy,'' and probabilistic reasoning; I.2.9 [Robotics]

General Terms:Wireless Sensor Networks, Robotics, IEEE 802.15.4

Keywords: Wireless sensor network (WSN), Locationidentification system (LIS), ZigBee, ANFIS, Vision robot

Received: 18 September 2014, Revised 28 October 2014,Accepted 6 November 2014

1. Introduction

The purpose of this paper is to construct an adaptive net-work based fuzzy inference system (ANFIS) [1] basedwireless sensor network (WSN) of location identificationsystem (LIS), constituted by the gateway, location nodes,server, client monitor, controller unit, etc. By using theseadvantages of low cost, low power consumption, a varietyof perceived control network connected to the network,concatenated into a sensor network through the networkcan achieve a variety of control techniques. The WSN isequipped with some radio transceivers and receivers toachieve information around the environment. When manyWSNs are deployed in large field, they can be automati-cally organized to form an ad hoc network [2, 3] to com-municate with each other by means of some network to-pologies such as star, mesh and tree communication to-pologies. In the past, many applications have been devel-oped such as vehicle/building automation, home secu-rity, environmental monitoring, indoor location identifica-tion, etc. These achievements let human life more com-fortable and convenient [4].

The WSN is envisaged to monitor the environment for manyyears. A challenge is to reduce energy consumption soas to extend the lifetime of the WSN. The IEEE 802.15.4/ZigBee Alliance is an association of companies workingtogether to develop standards (and products) for reliable,cost-effective, low-power wireless networking. The ZigBeetechnology will probably be embedded in a wide range ofproducts and applications across consumer, commercial,industrial and government markets worldwide [5, 6]. The

Journal of Digital Information Management Volume 13 Number 1 February 2015 37

ZigBee builds upon the IEEE 802.15.4 standard whichdefines the physical and medium access control (MAC)layers for low cost, low rate personal area networks. TheZigBee defines the network layer specifications for star,tree and peer-to-peer network topologies and provides aframework for application programming in the applicationlayer [5-9].

The applications of indoor LIS along with the popularity ofGlobal Position System (GPS) have been widely acceptedby consumers. However, the GPS signal is susceptibleto building shelter such as unable to provide indoorlocation identification services, for example, in small spaceindoor environment; the accuracy of GPS is unable tomeet the various needs. In addition to GPS, we can useLIS to achieve the purpose of positioning [10]. Manylocalization algorithms first use a ranging technique toestimate the Euclidean distances between nodes, andthen use the least-squares trilateration (LST) algorithm todetermine the locations of sensor nodes by theseestimated distances [11]. LIS based localization is one ofthe most prolific themes since radio signal reception isthe main requirement for its exploitation. Moreover, in thecase where the mobile receives frames from beacons andprocesses them to extract its position, there is no multi-user interference problem to consider such as two nodesreceiving data do not disturb each other [12].

For the analysis of vision robot studies have been growingrecently due to great interest among people. Some topicswhich can be reviewed are: evaluation of drinking waterquality, driving fatigue, health risks in work environment interms of injury, illnesses, safety, evaluation for healthpolicy, security issues among hospitals and its feasiblesolutions, and health services management. The firstapplication of fuzzy logic in health care was thedevelopment of an index to assess the health of thepatient. Another expert system was designed to helpphysicians and hospital staff in administrative, diagnostic,therapeutic, statistical, and scientific work. In this system,there are separate data-storing, health insurancesupporting, and simple advisory programs [13]. Recently,many advisory or decision monitoring programs have beendeveloped; for examples, by using neural networks to planfor monitoring patients in hospital environments [14]. Inthis paper, the neural network of ANFIS is used to combinewith the LIS to develop application of vision robot formonitoring the positions of patients.

The most useful property of neural networks is their abilityto approximate linear or nonlinear mapping throughlearning. Based on this property, the neural network basedcontrollers have been developed to compensate the effectsof nonlinearities and system uncertainties, so that thestability, convergence and robustness of the controlsystem can be improved [15]. The ANFIS [1] has beenproposed many years ago and widely used in researchworks. The ANFIS reveals an efficient learning networkand its applications can be found in [16, 17]. In this paper,the ANFIS controller is used to control the motors of fans.

The ANFIS controller is a suitable off-line hybrid learningprocedure to serve as a basis for constructing a set offuzzy if-then rules with appropriate membership functionsto generate the stipulated fuzzy associated memory (FAM)input-output pairs and fuzzy rules. Then a fuzzy inferencesystem (FIS) matrices can be achieved. By using thisFIS matrices, a reliable controller can be designed. TheMATLAB™ has provided very useful and friendly tools forengineers to design this ANFIS controller.

In this paper, the good performances of position monitoringof LIS are possessed by this proposed ANFIS based LISvision robot system. This method has also great benefitsfor reducing many burdens of nurses or doctors for caringpatients’ positions by using ANFIS based LIS vision robot.

1.1 Introduction to Location Identification System(LIS)The ZigBee Alliance is an association of companiesworking together to develop standards (and products) forreliable, cost-effective, low-power wireless networking.ZigBee technology will probably be embedded in a widerange of products and applications across consumer,commercial, industrial and government marketsworldwide. ZigBee builds upon the IEEE 802.15.4 standardwhich defines the physical and Medium Access Control(MAC) layers for low cost, low rate personal area networks.ZigBee defines the network layer specifications for star,tree and peer-to-peer network topologies and provides aframework for application programming in the applicationlayer [7]. In this paper, the subroutines are develop for thevision robot application. All these subroutines includedifferent type of application program interface (API). Forexample, the Queue API provides a queue-based interfacebetween an applications and both the IEEE 802.15.4 stackand the peripheral hardware drivers.

A variety of network topologies are possible with IEEE802.15.4. A network must consist of a minimum of twodevices, of which one Co-ordinator, referred to as thepersonal area network (PAN) Co-ordinator. The possiblenetwork topologies are Star topology, Tree topology andMesh topology. The basic type of network topology is theStar topology. A Star topology consists of a central PANCo-ordinator surrounded by the other nodes of the network,often referred to as End Devices. The Tree networktopology has an implicit structure based on parent-childrelationships. Each node has a parent. The node mayalso (but not necessarily) have one or more children. Eachnode can communicate only with its parent and its children.Any node which is a parent acts as a local Co-ordinatorfor its children. In the Mesh network topology, all devicescan be identical and are deployed in an ad hocarrangement. Some nodes can communicate directly. Notall nodes may be within range of each other, but a messagecan be passed from one node to another until it reachesits final destination.

An indoor location identification system (LIS) considersonly indoor environments such as inside a building. The

38 Journal of Digital Information Management Volume 13 Number 1 February 2015

location of users or their devices in personal area networkscan be determined by LIS measuring the location of theirmobile devices in an indoor environment. From thisdefinition, LIS should work all the time unless the userturns off the system, offer updated position information ofthe target, estimate positions within a maximum timedelay, and cover the expected area the users require touse an LIS [14]. The mostly adopted approach of LIS is toestimate the distance of the mobile unit from some set ofmeasuring units, using the attenuation of emitted signalstrength. Signal attenuation-based method such as LISattempts to calculate the signal path loss due topropagation. Theoretical and empirical models are usedto translate the difference between the transmitted signalstrength and the received signal strength into a rangeestimate [13].

The LIS used in this paper is called i-Tracer, it isdeveloped by least-squares trilateration (LST) algorithm[10]. LSI methods attempt to calculate the signal pathloss due to signal propagation attenuation. Fromtheoretical and empirical test model, the differencebetween the transmitted signal strength and thereceived signal strength can be translated into rangeestimation [13]. All data can be stored in database ofLIS by SQL and Java language programs. The serverarchitecture diagram of LIS is shown in Figure 1. It isconstructed by 3 servers to manage data of messageprocessing, position and monitoring [10]. These resultswill be shown in the following Experiment Result Sectionof this paper.

Figure 1. The Architecture Diagram of LIS

The hybrid learning algorithm developed in [1] can beapplied to Equation (3) directly. A neural network structureof ANFIS is shown in Figure 2. In the forward pass of thehybrid algorithm, functional signals go forward till layer 4

2. Preliminary of ANFIS

The ANFIS uses a hybrid learning algorithm to identifythe membership function parameters to generate Sugenotype fuzzy inference systems (FIS). It uses the methodof combination of least-squares and backpropagationgradient descent methods to train FIS membershipfunction parameters to model a given set of input/outputdata. The principle of ANFIS is briefly described as follows:[1]

Ri: if x1 is Ai1 ... and xn is Ain

thenui = pi1x1+.. +pinxn + ri (1)

where Ri denotes the ith fuzzy rules, i =1, 2, .., r; Aik is thefuzzy set in the antecedent associated with the kth inputvariable at the ith fuzzy rule; and pi1,...., pin , ri are the fuzzyconsequent parameters.

Based on the weighted averaged method of defuzzification.The output u can be calculated as

u =w1

w1+.. + wnu1+.. +

wn

w1+.. + wn

un

= w1u1+.... + w2un(2)

Because the fuzzy inference system is a Takagi-Sugenotype, i.e. ui = pi1x1 +.... + pinxn + ri, Equation (2) can berewritten as

where wi is the ith node output firing strength of the ith

rule; and

w1 =w1

w1+.. + wn ,..,. wn=

wn

w1+.. + wn

(3)

u = w1u1+.... + wn un

= (w1x1) pi1+.... + (w1x1) pin + (w1) r1+...

+ (wnx1) pi1+.... + (wnxn) pin + (wn) rn.

Journal of Digital Information Management Volume 13 Number 1 February 2015 39

of Figure 2 and the consequent parameters pi1, pi2, ri areidentified by the least squares estimate (LSE) approach.In the backward pass, the error rates propagate backwardand the premise parameters x1, x2 are updated by thegradient descent approach. As the values of theseparameters change, the membership functions varyaccordingly; thus exhibits various forms of membershipfunctions on linguistic labels Ai1 and Ai2.

3. Experimental Results

Consider a vision robot such as in Figure 3 is walking onO-XY plane such as the state space dynamic equationsare

x = Ax + Buy = Cx (4)

The program of WSN is developed on the free softwarecalled Code Blocks. At first, the program of Co-ordinatoris developed then the program of End Device is developedconsequently. Every network must have one and only onePAN Co-ordinator, and one of the tasks in setting up anetwork is to select and initialize this Co-ordinator. Thepersonal area network identification (PAN-ID) must be setadequately in program. This is a high power ZigBee Kit. Itcan provide all the software tools and hardware requiredto get the first-hand experience with WSN. The entry-level kits contain one base development board (BDB) andone sensor development board (SDB). Each board isequipped with a high-power IEEE 802.15.4 RF modulebased on LIS MCU which provides much higher coveringrange with 2.4GHz RF antenna which is with the IPEXconnector for easy mechanical design than normal-powerRF module. For I/O expansion ports, it has 10 useful pinsof GPIO include UART, ADC, DAC and Comparator.

For the software, it also provides free ApplicationProgramming Interface (API) packages to the peripheraldevices of single-chip IEEE 802.15.4 compliant wirelessmicrocontrollers. This is known as the IntegratedPeripherals API. It details the calls that may be madethrough the API in order to set up, control and respond toevents generated by the peripheral blocks, such as UART,GPIO lines and Timers among others. The softwareinvoked by this API is present in the on-chip ROM. ThisAPI does not include support for the ZigBee WSN MAChardware built into the device; this hardware is controlledusing the MAC software stack that is built into the on-chip ROM. ZigBee can be used with different sensorssuch as in vehicle or home automation, securitymanagement, industrial, environmental controls, andpersonal medical care. In this paper, the ZigBee WSNsare used to design for the vision robot by means of ANFISbased control methodology. At first, the LQI test is testedfrom the results of the values of LQI. It will be variedaccording to the distance between the Co-ordinator andEnd Device. All these data will be stored in database namedas file of RTLS.mdf, these data can be used by SQLprogram and can be feed to ANFIS system to construct a

The ANFIS based controller described as aforementionedin Section 2 is developed. If there are three location nodesand three tags placed in three corners. From the strengthdata, after ANFIS based LIS controller’s manipulations,the fuzzy values will be returned to the LIS server to monitorthe positions of these corners. The ANFIS design resultsare shown in Figure 4. The inference ANFIS output valueis normalized from 0 to 1 can be achieved. In this condition,for example, the node position is far, the ANFIS value willdown, and otherwise, its value will be high. If the value islower than the threshold value, then the vision robot willbe active to control the moving path. The ANFIS-basedfuzzy inference mechanism or fuzzy rules are developed,it just need 6 fuzzy rules. In this experiment, the data ofevery location node or tag are used. The ANFIS basedLIS for vision robot in this paper just illustrate an exampleof node position caring. From Figure 4(d), the vision robotwill reach to location node 1. There are many otherapplications can be developed by this LIS. From thisexample, it demonstrates that the ANFIS based LIS issuccessfully established, meanwhile, the good vision robotperformance is also possessed.

vision robot system.

Figure 2. A two-inputs one-output ANFIS architecture

Figure 3. LIS based diagram of vision robot

4. Conclusion

In this paper, the design method for the application ofvision robot by using the adaptive-network-based-fuzzy-inference-system (ANFIS) based wireless sensor network(WSN) location identification system (LIS) is proposed.

.

40 Journal of Digital Information Management Volume 13 Number 1 February 2015

Figure 4 (a) One of the ANFIS membership functions diagram

Figure 4 (b). One of the ANFIS inference surfaces diagram

Figure 4 (c). The fuzzy rules of ANFIS

Journal of Digital Information Management Volume 13 Number 1 February 2015 41

Figure 4 (d). The inference result of ANFIS

This paper has successfully demonstrated the applicationof the LIS to monitor the location of vision robot. Thephysical verifications and simulations are alsosuccessfully demonstrated that it has possessed the goodperformances of ANFIS based LIS for one example ofvision robot applications.

AcknowledgmentThe author appreciates the partial financial supported bythe Ministry of Science and Technology, Taiwan, R. O. C.under Grant of MOST 103-2221-E-253-003.

References

[1] Jang, J. S. R. (1993). ANFIS: adaptive-network-basedfuzzy inference system, IEEE Transactions on Systems,Man and Cybernetics, 23, p. 665-685.

[2] Hu, Y. C., Johnson, D. B., Perrig, A. (2003). SEAD:secure efficient distance vector routing for mobile wirelessad hoc networks, Ad Hoc Networks, 1, p. 175-192.

[3] Abolhasan, M., Wysocki, T., Dutkiewicz, E. (2004). Areview of routing protocols for mobile ad hoc networks,Ad Hoc Networks, 2, p. 1-22.

[4] Diallo, O., Rodrigues, J. J. P. C., Sene, M. (2012).Real-time data management on wireless sensor networks:A survey, Journal of Network and Computer Applications,35, p. 1013-1021.

[5] Golmie, N., Matta, I. (2005). Applications and servicesin wireless networks, Computer Communications, V. 28,p. 1603-1064.

[6] Su, I. J., Tsai, C. C., Sung, W. T. (2012). Areatemperature system monitoring and computing based onadaptive fuzzy logic in wireless sensor networks, AppliedSoft Computing,12, p. 1532-1541.

[7] Baronti, P., Pillai, P., Chook, V. W. C., Chessa, S.,Gotta, A., Hu, Y. F. (2007). Wireless sensor networks: Asurveyon the state of the art and the 802.15.4 and ZigBeestandars, Computer Communications, V. 30, p. 1655-

1695.[8] Yick, J., Mukherjee, B., Ghosal, D. (2008). Wirelesssensor network survey, Computer Networks, 52, p. 2292-2330.[9] Aguilar, L., Licea, G., García-Macías, J. A. ( 2011). Anexperimental wireless sensor network applied inengineering courses, Computer Applications inEngineering Education, 19, p. 777-786.[10] i-Tracer User Guide, Fontal Technology Inc., 2010.(http://surewin.com.tw)[11]Lu, K., Xiang, X., Zhang, D., Mao, R., Feng, Y.Localization algorithm based on maximum a posteriori inwireless sensor networks, International Journal ofDistributed Sensor Networks, Article ID 260302, 2012.(doi:10.1155/2012/260302)[12] Dalce, R., Val, T., Bossche, A. V. D. (2011).Comparison of indoor localization systems based onwireless communications, Wireless Engineering andTechnology, 2, p. 240-256.[13] Abbod, M. F., Keyserlingk, D. G., Linkens, D. A.Mahfouf, M. (2001). Survey of utilisation of fuzzytechnology in medicine and healthcare, Fuzzy Sets andSystems, 120, p. 331-349.[14] Bajo, J., Paz, J. F., Paz, Y., Corchado, J. M. (2009).Integrating case-based planning and RPTW neuralnetworks to construct an intelligent environment for healthcare, Expert Systems with Applications, V. 36, p. 5844-5858.[15] Mon, Y. J. (2012). Supervisory recurrent fuzzy neuralnetwork control for disturbed nonlinear system, AdvancedScience, Engineering and Medicine, 4, p. 350-354.[16] Mon, Y. J. (2007). Airbag controller designed byAdaptive-Network-Based Fuzzy Inference System(ANFIS), Fuzzy Sets and Systems, V. 158, p. 2706-2714.[17] Buragohain M., Mahanta, C. (2008). A novel approachfor ANFIS modelling based on full factorial design, AppliedSoft Computing, 8, p. 609-625.