2010 2nd International Coference on Computer Technoloy and Development ICCTD 2010) Improving Directed Diffusion in Sensor Network Using Learning Automata:DDLA New Approach in Directed Diffusion Ali Sayyad Msc. Information Technology, Network Eng. Computer & Elec. Department Islamic Azad University of Qazvin Qazvin,Iran later _ opt@yahoo.com Mohammad Shojafar Msc. Computer Eng. Computer & Elec. Department Islamic Azad University of Qazvin Qazvin,lran Shojafar@qiau.ac.ir Abstract-O ne of the important and challenging matters in sensor network is energy of life span of nodes in the network. Node's movement, speciically movement of central node (sink) in these networks cause to increase routing updating overhead and consequently to increase power consumption and to decrease network life span. Directed Difusion algorithm is one of methods in sensor network which is a data-oriented algorithm. One of the important deinitions of basic algorithm is not supporting central node's movement. In case of movement of central node, data packs pass on unreliable rout toward central node. In fact, they pass a rout on which the central node is not present at the time being and it has moved to another place. Therefore, route of data is out of order and there is the need to build new routes. This problem causes to create lots of overhead and waste energy. In this article, it is tried to solve mentioned problem of central node's movement by learning automata. In suggested algorithm by learning automata a route amendment tree is built which prevents from creation of the whole route and its overhead. Keywords-component; Wireless Sensor Network; Directed Dffusion; Learning Automata; Dynamics I. INTRODUCTION Current progress in technology of short radio range and micro electro mechanical system leads in production intelligent sensors which are generally calculated by capability of data sense, wireless connection in device is strongly limited with regards to equipped resources. These tiny sensors can be considered for gathering information and they can be distributed to central controlling units named sink in diferent regions. For transferring sensed data in the network to sink, many wireless steps may be taken through a collection of small and limited sensor nodes with regards to energy. This characteristic is important for difusing data in wireless sensor network. Within previous years, diferent algorithms and protocols were suggested with goal of achieving more eficiency and 978-1-4244-8845-2/lO/ $ 26.00 © 20lO IEEE 189 ZiaDelkhah Msc. Information Technology, Network Eng. Computer & Elec. Department Islamic Azad University of Qazvin Qazvin,Iran ziadelkhah@gmail.com Mohammad Reza Meybodi PHD in Computer science. Computer & IT. Department Amirkabir University,Tehran,Iran mmeybodi@aut.ac.ir reliabiliy for diusing data in wireless sensor networks. For example, looding difusion is the most reliable method for sending data rom sensor to a sink. This protocol doesn't impose any extra cost for keeping topology and discovery of the route and its implementation is very simple and easy. But, main problem of this protocol is creating too much overhead in the network due to sending repeated messages. This problem specially leads to ineiciency in energy consumption. A new data-oriented method which was different rom traditional routing was created for wireless sensor networks in order to solve this problem. The main goal for generating this method is because of concentrating of sensor network on the data generated with sensors and it doesn't focus on sensed nodes. Data-oriented protocols are based on query and are dependent on naming (for instance mentioning data with binary of trait-value) of desirable data. Therefore, too much transfer of many packs is prevented and it causes energy eiciency. Direct difusion[l] is an instance of one of data-oriented protocols because of its characteristics for saving energy, such as, data query based on query (named interest) by sinks, collecting and saving data by sensor and mechanism of gradient and route enhancement. Gradient concept usually means direction toward those neighboring by which sink is accessible. In wireless sensor network, most data packs are sent toward sink rom a sensor complex. Therefore, here, each sensor node's important task is to create and keep gradient value in each node. Totally gradient value management is done by primary and requent primitive looding diusion of a series of a controlling pack (like interest packs in direct difusion [1]) rom a sink. Please note that this requent looding diffusion all over the network leads in too much overhead (regarding band width and energy consumption) in sensor networks. Moreover, when network topology change because of failure of senor nodes, wireless connections and value of some gradients will become unreliable, thus, there is need to requent looding diusion. This protocol is sensitive to node's movement due to direct diusion nature. In fact, the best dynamism plan should establish a minimum