A novel deep neural networks based path prediction
- PDF / 720,878 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 95 Downloads / 228 Views
(0123456789().,-volV)(0123456789(). ,- volV)
A novel deep neural networks based path prediction Upasna Joshi1 • Rajiv Kumar1 Received: 24 November 2019 / Revised: 24 November 2019 / Accepted: 23 January 2020 Ó Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract With the advancements in sensor applications, wireless sensor networks (WSNs) become significant area of research. WSNs compose various tiny sensor nodes to sense an environment, depends upon the given application. However, these nodes are battery constrained (i.e., may become dead after passing certain iterations). Therefore, number of energy efficient protocols have been implemented in literature. However, selecting an optimal path between base station and sensor nodes is defined as an ill-posed problem. To overcome this issue, a Deep Q-routing based inter-cluster data aggregation is considered to improve the inter-cluster communication in WSNs. In every epoch, Recurrent neural network is considered to compute shortest path between cluster heads and sink. We have trained the network in such a way that it considers various features of WSNs and able to decide which sensor node will be selected as next-hop to establish a shortest path between elected cluster heads and sink. The non-cluster head nodes may also be considered while selecting a shortest path. Extensive experimental results show that the proposed technique outperforms the competitive energy efficient protocols. Keywords Energy efficiency Q-learning Deep learning Non-dominated solutions
1 Introduction Wireless sensor network (WSNs) is a latest area of research which has wide number of applications in the field of computing, medicines, armed forces etc. [1]. It consists of large number of sensor nodes which are randomly distributed in order to monitor physical and environmental situations [2]. Latest research in this area shows that many researchers are planning to develop different novel conventions which will especially focus on energy constraint [3]. Generally, the concentration has been redirected towards routing procedures as they may vary based on its network design and vast applications [4]. Leu et al. [5] used Regional Energy Aware Clustering with Isolated Nodes (REAC-IN) to increase the lifetime of WSNs. In this, isolated nodes send the data to CH node based on the regional average energy. Pal et al. [6] improved the lifetime of WSNs by implementing the
& Upasna Joshi [email protected] 1
Computer Science and Engineering Department, Thapar Institute of Engineering and Technology, Patiala, Punjab, India
balanced cluster size. The thresholds of cluster formation are used for even distribution of clusters. Chidean et al. [7] used Data-coupled clustering (DCC) to enhance the efficiency of WSNs. In this, the combination of clustering and in-network processing techniques are utilized. Lee and Kao [8] implemented the semi-distributed clustering technique based on three-layer hierarchy to increase the efficiency of WSNs. Alami and Najid [9] used an Improved
Data Loading...
