Research On Distributed Parallel Multiobjective Evolutionary Algorithms And Their Application Research On Wireless Sensor Network Deployment

In real life,many kinds of problems,especially NP-hard problems,can be tackled by utilizing evolutionary algorithms.When the optimization problems involve several generally conflicting objectives,they can be regarded as multiobjective optimization problems,which can be optimized by employing multiobjective evolutionary algorithms.With the advent of the big data era,more and more multiobjective large-scale optimization problems are emerging.Traditional multiobjective large-scale evolutionary algorithms are almost serial,so novel parallel algorithms are needed.Multiobjective evolutionary algorithms can be applied to wireless sensor network deployment problems.Traditional research focuses on 2D plane and 3D full space,and the sensors are mostly homogeneous omni-directional.However,this cannot meet the assorted demand of real-world complex 3D environments.Therefore,this paper employs evolutionary algorithms to optimize multiobjective deployment problems of wireless sensor networks in complex 3D environments.This paper is composed of four parts.First,a distributed parallel cooperative coevolutionary multiobjective evolutionary algorithm is proposed.On the basis of an improved grouping method,variables are decomposed,and each group is optimized by a subpopulation,which is further divided into multiple sets;accordingly,a two-tier parallel structure is constructed.This algorithm decomposes the original problem into several simpler small-scale problems,which decreases solving difficulty;additionally,parallelism significantly reduces optimization time.Second,the deployment problem of a heterogeneous wireless sensor network on 3D terrains is studied.We propose an improved uncertain coverage model to make it more suitable for practical application.Coverage,connectivity uniformity and deployment cost are simultaneously considered,and the multiobjective evolutionary algorithms are utilized for optimization.Third,a distributed parallel cooperative coevolutionary multiobjective large-scale immune algorithm is presented.Multiple populations can sufficiently explore each objective,and the large-scale problem can be decomposed into multiple small-scale problems through variable grouping.Introducing the idea of the immune algorithm makes the algorithm prefer exploring the sparse regions in the objective space and improves uniformity and diversity of the obtained solution set.The application to the multiobjective deployment problem discussed in the second part validates the presented algorithm's effectiveness.Finally,the algorithm in the first part is improved: at the expense of partial parallelism loss,the individual evolution process is performed in serial in a single CPU,thus,the information of the whole subpopulation can be comprehensively exploited.We have experimented various optimizers,and verified the good expansibility of this distributed parallel structure.The deployment research of industrial wireless sensor networks in the complex 3D engine room space with obstacles has been conducted,and an improved uncertain coverage model is put forward.The reliability is regarded as one of optimization objectives,and coverage,lifetime and reliability are comprehensively taken into consideration.Multiobjective evolutionary algorithms are utilized for optimization,and the experimental results have confirmed the improved algorithm's superiority.In summary,this paper proposes several distributed parallel cooperative coevolutionary multiobjective large-scale evolutionary algorithms.The multiobjective deployment problems of heterogeneous directional sensor networks on 3D terrains as well as industrial wireless sensor networks in the 3D engine room space with obstacles are studied,the application of the presented algorithms to which further validates their effectiveness.