Research On Adaptive Task Allocation Algorithm In Wireless Sensor Network

Wireless sensor networks (WSNs) is usually made up of a large number of sensornodes whose ability of storage is very limited as well as computing and communication,and these nodes are connected through the way of self-organizing and wireless to formthe network. Recently, people have done relatively deep research on wireless sensornetworks, but it still has considerable research space for the problem of adaptive taskallocation in sensor network. As wireless sensor networks have several characteristics,as follows: the instability of its load; the limit of the resources and executing ability ofthe nodes; the complexity and real-time characteristic of the application, so, how tocomplete the tasks in wireless sensor network effectively and accurately is an importantproblem which the design of the algorithm of the networks need to consider.This paper will do a comprehensive research on the adaptive task allocationproblem of wireless sensor networks. The main work includes three phases, as follows:(1)Considering the limit of the resources and large redundancy of the network, thepaper introduces the theory of dynamic alliance in order to avoid unnecessaryconsumption of the resource. Propose a serial alliance based dynamic task allocationalgorithm in wireless sensor network, and this algorithm is mainly used to solverelatively simple tasks or some applications whose real-time characteristic is weaker.The paper develops a discrete particle swarm optimization (DPSO) based dynamicalliance generating algorithm, and the algorithm enables to an optimal executingalliance for each task; then, the paper also applies over-demanding algorithm to allocateeach task to the members of corresponding alliance. Besides, the proposed algorithmwill track the status of the network in real-time and do necessary adjustment.(2)Due to the characteristics of real-time and uncertainty in wireless sensornetworks, the paper proposes a parallel alliance based dynamic task allocation.Through designing a DPSO based parallel alliance generating model, the network cangenerate some suitable alliance for several unassigned tasks in parallel; then, in order toimprove the comprehensive performance of parallel alliance, the paper also appliesover-demanding algorithm to allocate the subtasks of each unassigned to their alliancesrespectively. In addition, the algorithm also adjusts the status of nodes when the statusof the network changes. Therefore, the lifetime of the network will be prolonged.(3)In order to thoroughly improve the property of the network, introduce thethought of game theory. The paper develops a game theory based adaptive task allocation. After choosing the corresponding alliance for each unassigned task, thesetasks will be allocated to there alliance respectively through a game theory basedallocation algorithm, so the performance of each alliance will be improved; at the sametime, for the confliction problem of resource in the process of allocating, the paperdesigns a conflicting eliminating algorithm applying game theory. Being different fromprevious algorithm, this algorithm synthetically considers the whole performance ofparallel alliances and each alliance of parallel alliances, so effectively prolonging thelifetime of the whole network.