Fuzzy Calculating and Fuzzy Control in Wireless Sensor Network

At present time configurable wireless sensor networks are given special consideration. Wireless sensor networks are now a static elements union. To make such a network a powerful system that is able to replace traditional computers, research is required of the application of different theories and technologies for sensor network development. However, building models of approximate reasoning and its usage in computer systems presents an important scientific problem. Fuzzy logic provides effective representation of real world information. The mathematical method of fuzzy information representation allows the building an adequate model of the reality. The article describes an algorithm of a fuzzy controller used to make soft computing and soft managements of a wireless network. Such a network consists of devices with limited resources which are able to work during many months and possibly years. It has been suggested that AI methods can be used to benefit the technological development of wireless sensor networks (WSN). Now, WSN conception changes the human role, as its elements sensor microcontrollers become more independent and underlying of human tasks. The Homocentric model of network calculating, with humans as the main network unit becomes a thing of the past. Humans shift from calculating central to its periphery, concentrating on process control, becoming the mediator between real world and computers. Actually a new class of distributed computers has been created which will open new outlooks. WSN has made its first step on the road to the next epoch when computers will be connected with the physical world. At this point they will be able to pre-determine users wishes, and will be able to make independent decisions. In last time, fuzzy controllers are widely adopted for practical issues. They allow managing of complex and ill formalized processes basing on linguistic information. In this paper, research of fuzzy logic algorithm application has been carried out for optimization of WSM work. One of the application examples can be optimization device work frequency. WSN consist of tiny devices with limited resources but they have to work without recharging over periods of many years. So issues of energy saving occur. To prolong the devices "life" methods of fuzzy analyses are suggested for more flexible management of device work. For example, if the network is intended for environment observation and has to report about unusual cases then the fuzzy controller algorithm is convenient to optimize device work frequency. If the situation around is normal, network nodes work less frequently to lose less energy. As soon as the situation oversteps the limits network the device begins to work more frequently to store all information. Stored data can be used to analyze by device itself, network or network administrator. Fuzzy controller block is placed in every network device. In our example, frequency of sensor data reading, the Fuzzy controller input is sensor data and output is a required value. If we take for example a network which watches the temperature conditions i.e. the network device periodically reads the temperature value. In this case the fuzzy control input is temperature value and the output is the temperature reading frequency value. A Network device containing such a fuzzy controller module can react with more flexibility to changes in its environment. This makes it possible to save a great deal of energy. This module works without the assistance of human interaction and so therefore provides a device for independency. Fuzzy controller input can be heterogeneous data, for example temperature and light or temperature and humidity. Thus a device can analyze different parameters depending on the task at hand. There are 2 main methods to design a fuzzy controller module to application work based on hardware capabilities. The first method is the use of a special device for fuzzy management applications. They are divided into 2 types: • Fuzzy coprocessors. These are special devices for execution of input fuzzy operations. They are controlled by universal microprocessors or microcontrollers, for example VY86C570 (Togai InfraLogic, Inc); • Microcontrollers, which have special commands and registers to execute fuzzy input, for example Motorola 68HC12; The Second method is the use of universal microprocessors or microcontrollers. Fuzzy controller module is implemented as software in this case. To do it there are several variants: • Using of fuzzy frameworks. They collect variables information, fuzzy sets and fuzzy operators which are selected by the user and generate program code afterwards; • Using of finished libraries. They are designed especially to develop fuzzy controller applications; • Developer designs whole system. Such a fuzzy controller module has a slower response rate than special fuzzy hardware. In spite of this, this particular way of fuzzy module development is more flexible in making some changes in fuzzy application. Such fuzzy controllers allow humans to try several various variants of applications without the devices changing. To design the fuzzy controller module it is necessary to create a Knowledge Base which should include concept information divided into Data Base and Rules Base. Knowledge Base is a static data structure which has predefined size and installed contents. Knowledge Base structure should be defined before starting application work. Data Base contains information about fuzzy sets which are linguistic terms used in fuzzy rules. To simplify a calculation, each fuzzy set membership function is presented as a piecewise-linear function and has three or four points. Points count depends on the membership function form: triangular or trapezoidal. These point values in a fuzzy set match up places which they have in variable definitional domain. An Example of variable definitional domain division into five trapezoidal fuzzy sets is shown on the picture below. Rules Base is rule set. Each rule uses matched fuzzy set from the Data Base. There are two possible data structure designs to store Knowledge Base. The decision depends on hardware capability. If computer is power optimal, structure design can consist of entries array where each entry is a rule from Rules Base and has variable fields (antecedents and consequents) of integer or point type. These integers or points are reference to fuzzy sets array. Sometimes it is necessary to use dynamic count of rules and fuzzy sets for several Knowledge Bases. In this case, static arrays should be replaced by lists in dynamic memory. Microcontrollers can use only simple data structures consisting of integer or float arrays. Structure bases on as many arrays as antecedents and consequents and defining points of fuzzy sets. For example for trapezoidal fuzzy sets Data Base is float arrays. Each element of these arrays contains information about antecedent i, fuzzy set j and defining fuzzy set point k. Rules Base consist of three arrays. These arrays size are rules count. Rules link to fuzzy sets which Data Base keeps. Each node of network has a fuzzy controller module which has a set of input and output linguistic variables. Thus this type of sensor node can be called a fuzzy sensor. Fuzzy sensor contains linguistic variables called also attributes, Knowledge Base consisted of fuzzy rules and машина вывода which carries out calculation of output variables. In order for the fuzzy system to be more flexible, capability of linguistic rules and Knowledge Base remote setting has been added. Thus it is possible for the fuzzy sensor to change its behavior if the user wishes, whether it is human or a computer program. Input variables are linguistic mapping of physical sensors placed on the device. Output variables values are results of fuzzy controller work. A set of such fuzzy sensors which are wireless network nodes can organize hierarchy. Thus, it is possible to form a distributed system of information processing and storing. The hierarchy is created in the following way, output variables of some fuzzy sensor can be input variables of other fuzzy sensor or sensors. Multilevel hierarchies of fuzzy sensors are created using this principle. This distributed system allows tasks of any complexity to be implemented. There is capability to build distributed Data Base to process fuzzy queries in wireless sensor network. This capability is easily realized, but is not expedient, as queries to Data Base and fuzzy modifiers presented in the form of ordinary relational Data Base are limited. Also the Data Base has a large memory size. Therefore the use of SQL extended to fuzzy case to process fuzzy queries is not effective. Following this a special fuzzy query language, and communications in distributed systems were developed. This language can be used for • sensor network requests • fuzzy triggers creating • fuzzy active Data Bases creating • fuzzy sensors communications It is based on this approach that the distributed Data Base was created. Its data is stored in various network nodes. The Client uses such distributed fuzzy Data Base as ordinary Data Base. The query can be both simple, for example a query of attribute values for a given period, and complex containing aggregation functions. Also it can be fuzzy which asks output attributes fuzzy calculation. If the node does not have enough information to execute a query it creates and sends its own queries to other nodes. Five general, goal-oriented, data fusion methods are in use today in WSN (ordered by data complexity) data association, identity fusion, effect estimation, pattern recognition and artificial intelligence. Ten discrete data fusion techniques can be identified within these five general categories: figure of merit and gating technique in the data association, Kalman filters in the identity fusion, Bayesian decision theory and Dempster-Shafer evidentional reasoning in the effect estimation, adaptive neural networks and cluster methods in the pattern recognition, expert systems, blackboard architecture and fuzzy logic in the artificial intelligence. The sensor data fusion technology focuses on the acquisition of high-level information (artificial intelligence level), i.e. information that is related to many conventional physical quantities in a nonanalytical way. In these complex cases, fuzzy production systems and fuzzy neural networks are more effective and they compute and report linguistic assessments of numerically acquired values. Two methods are proposed to realize the aggregation from basic measurements. The first one performs a combination of the relevant features by means of a rule-based description of the relations between them. With the second, the aggregation is realized through an interpolation mechanism that creates a fuzzy partition of the numeric multi-dimensional space of the basic features. This partition can be realized with fuzzy neural networks. But fuzzy sensor can also can be used on low levels of data fusion, e.g. for filtering and for pattern recognition. Aggregation functions can be modeled using T-norm and T-conorm. To do it only one membership function of line form should be built. But rules are different for various aggregation functions. For example it is required to calculate maximum of two temperature values which are measured different nodes. Two linguistic variables, Temp1 and Temp2, should be created and membership functions, f1 and f2, should be built for these variables. So every variable has one membership variable of line form. Result is output linguistic variable Max which has line function fmax as Max membership function. In this case rules have to be following: IF Temp1 =f1 OR Temp2 = f2 THEN Max =fmax To calculate minimum of Temp1 and Temp2 rules have to be following: IF Temp1 = line AND Temp2 = line THEN Max = line Factors which can influence on exactness of fuzzy logic system has to be take into account. For example: • selecting of fuzzy sets kind and number • selecting of defuzzyfication method • selecting of operators which are used in rules. Operators selecting is very important and directly influences on result exactness. In this fuzzy system following T-norm and T_conorm are used: • Zade’s T-norm and T_conorm • probabilistic T-norm and T_conorm • Lukasevich’s T-norm and T_conorm • parametric Franc’s nd T-conorm • parametric Sugeno’s d T-conorm The same system can use various types of Tnorm and T-conorm. Setting can be made in any moment during system working. Such module can be separate into following parts: • fuzzy controller algorithm extended to using of T-norm and T-conorm • initial data download