Path management method using partially connected neural network in large-scale heterogeneous sensor network

초록

In this paper, we develop the cost function of the path management method for data delivery in large-scale heterogeneous sensor network. Usually, the most conventional methods determine the optimal coefficients in the cost function, without considering the node surrounding environments, such as the wireless propagation environment or the topological environment. Due to this reason, there is the limitation to improve the performance of path management, such as data delivery ratio and delay of data delivery. To solve this problem, we derive a new cost function using the concept of partially connected neural network (PCNN) that is modeled according to the input types whether inputs are correlated or uncorrelated. In our application, we assume that all inputs of the cost function are uncorrelated. Thus, we connect all inputs to the hidden layer of the PCNN in an uncoupled way. We also propose the training technique for finding the optimal weights in the PCNN. Our PCNN is trained to maximize the packet transmission success ratio. In the experimental section, we show that our PCNN method outperforms other conventional methods in terms of the quality of data delivery, such as data delivery ratio and delay of data delivery.