Learning Automata for Wireless Sensor Networks

Abstract

Wireless senor networks (WSN) are collections of tiny and low-cost sensors, capable of wirelessly communicating and cooperating with each other to fulfill intended network duties. The environment of a WSN is highly dynamic due to a number of reasons, topological changes, wireless errors, autonomity of sensor nodes which results in ad hoc changes in their behaviors, resource limitations, and so on. Considering such dynamism, it is evident that the algorithms and protocols designed for WSNs should be made adaptive so that they can cope with the environmental changes occur in such networks. Considering the above mentioned characteristics of WSNs, this chapter aimed at justifying the suitability of the learning automaton (LA) model in designing algorithms and protocols for WSNs. There are strong evidences which support this idea. First, LA is proved to perform well in distributed environments like the environments of WSNs, where the number of distributed elements is very large and the overhead of using centralized algorithms is very high. Second, LA has a very low computational and communicational overhead which makes it an outstanding model to be used in resource limited environments as of WSNs. Third, LA model is highly adaptive to the environmental changes, and hence, is well-suited to highly dynamic environments like the environments of WSNs. Finally, the reinforcement signal used by the LA is considered as a random variable and hence, its instant values do not affect the performance of the LA in the long run. To justify the idea, a number of learning automata-based algorithms will be proposed for addressing five different problems within the area of WSNs; data aggregation, clustering, deployment, k-coverage, and dynamic point coverage. For any of the proposed algorithms, extensive simulation studies will be reported to elaborate the performance of the algorithm. Results of these studies strongly support the idea that LA is a suitable problem solving model for the area of WSNs.