Research On Routing Design Of In-network Processing In Wireless Sensor Network

In many monitoring applications of wireless sensor networks?WSNs?,such as structural health monitoring,volcano tomography,and machine diagnosis,the raw data in WSNs are required to be losslessly gathered to the sink,where a specialized centralized algorithm is then executed to extract some global features or model parameters.To reduce the large raw data transmission,in-network processing is usually employed.In-network processing means computing and transmitting partially aggregated data rather than transmitting raw data in networks.However,most of the existing in-network processing work often makes some general assumptions about the calculation functions?also called fusion functions?on each sensor node,and does not pay attention to whether the calculations performed on these nodes are consistent with centralized processing tasks of the specific monitoring application.This paper studies how to perform sub-tasks on different sensor nodes for a given centralized processing task,so that the final calculation results are consistent with a given centralized calculation.We call this lossless in-network processing.It can be seen that the lossless in-network processing requires equivalent partition of the given centralized algorithm,and then the partitioned computing is reasonably allocated to different nodes in the network.At present,these problems have not been sufficiently studied in the industry.This paper studies the two problems around the lossless in-network processing: 1)what pattern should a centralized algorithm be partitioned into so that the partitioned computations can be flexibly assigned into a WSN with arbitrary topology? And 2)for each partition pattern,how should efficient routing for the resource-limited sensor nodes be designed? These two questions can be called as a topology-constrained computation partition problem and a computation-constrained routing design problem,respectively.This paper first introduces a general partition pattern suitable for any network topology,and then discusses a series of new routing methods different from the traditional routing form for various partition pattern.For the partition pattern of nested tree,we design specific routing algorithms under different calculation rules.Especially for the rule of linearly increasing,We prove that the data transmission cost of the proposed routing algorithm does not exceed 1+2^1/2 times the cost of the optimal intra-process tree.In the thesis,a lot of simulations are carried out on the proposed algorithm.The simulation results show that the proposed routing algorithm can significantly reduce the transmission overhead compared with the classical routing algorithm under the same calculation results.