Research On Dynamic Nodes Deployment Of Hybrid Wireless Sensor Networks With Grid Partitioning

An important research direction in wireless sensor network(WSN) is the study of hybrid WSN, i.e., WSN with dynamic nodes, including increase of coverage rate, reduction of redundancy rate, attack detection, and small-world network etc. While many previous reports involved complex application of node mobility, very few focused on the path planning and scheduling of dynamic nodes in hybrid network with many dynamic nodes. With the development and utilization of big data theory, research on deploying dynamic nodes in large-scale hybrid WSN will be a hot spot. In the present study, we explored the path planning and rescheduling of dynamic nodes in respond to emergencies in large- and super-scale hybrid WSN.General scheduling schemes for dynamic nodes in hybrid WSN only consider how to reduce the energy consumption. However, besides the minimization of energy consumption, its balance is also an important factor that should to be paid attention to. Imbalanced energy consumption will induce untimely ends of some dynamic nodes, thus increase the work load of the other dynamic nodes, resulting in the shortening of dynamic nodes lifespan. In this study, we proposed a energy balance model to schedule dynamic nodes according to multiple traveling salesman problem(MTSP) scheme with minimax constraints, which was solved by genetic algorithms; for some special cases, such as the locations and number of events are equivalent to those of dynamic nodes, we used a bipartite graph matching and clusters algorithms for dynamic nodes scheduling to reduce the time complexity. Combining both approaches, we obtained the final energy balance centralized synthesized algorithm.For our present centralized synthesized algorithm as well as others previously published, a central server is necessary to collect and process information from the whole network. In a large-scale hybrid network, if all nodes are simultaneously transmitted to the base station for data processing, a huge amount of data packet forwarding will be produced as the nodes information can only be transmitted by node-to-node. In addition, even in the static network, the "funnel effect" can be generated by the huge scale of network. A considerable solution is to divide the large-scale network into grids, and implement the autonomy of the centralized algorithm in each grid. However, the randomicity of events and uncertainty of node locations make it very difficult to carry out. In this study, we designed a hexagonal grid partition model, and developed a length-search message exchange mechanism based on the model, which greatly reduced the communication energy consumption. Combining this approach with existing methods of grid computing, we proposed a algorithm for distributing dynamic nodes in different grids. Finally, integrating the intra-grid and inter-grids scheduling approaches, we got a distributed algorithm based on hexagonal grid for dynamic nodes scheduling.In the simulation experiments, we compared the two different algorithms and explored their advantages and defects. We also verified the effectiveness of the proposed algorithms by comparing with other existing algorithms. At last, we summarized the limitations of present study and proposed for the future work.