Research On Traffic-aware Routing In Software-defined Wireless Sensor Networks

As the infrastructure of Internet of things(IoT),wireless sensor network(WSN)connects the physical world and the Internet world.With the development of IoT and smart city,sensor nodes are becoming more and more popular.Due to the limitation of sensor resources,the traditional wireless sensor network structure has been unable to meet the quality of service requirements of data communication.At the same time,WSN is faced with difficulties in expansion and inability to synchronize awareness.Software defined networks(SDN)bring opportunities for WSN,which have the characteristics of programmability,separation of data forwarding and control decision.We introduce SDN for WSN and design a software defined wireless sensor network(SDWSN)communication mode for large-scale wireless sensor nodes.The design fully considers the resource constraints of the nodes and QoS requirements during data transmission.We study the routing and topology control of sensor networks.The main work of this paper is as follows:Firstly,this thesis analyzes wireless sensor network and the possibility of application software definition.And we analyzed the different routing technologies of WSN in detail.Secondly,we designed the software to define the interaction process between the control plane and the data plane of WSN,which designed the traditional sink node as the controller and adopted the centralized control decision.The controller periodically generates data forwarding rules according to the view of wireless sensor network.Because the controller in SDN has a global view,it can better control the data forwarding of distributed nodes.The controller adopts periodic network state collection,so the overhead of control message is reduced by round.In this thesis,two different machine learning algorithms are applied to the fixed transmission range communication and the topology variable range communication in WSN,and solutions for different dynamic environments are proposed.We deeply study the way of deep learning to optimize routing.The development of deep learning provides a feasible way to optimize large-scale control problems,so machine learning algorithm is used in the controller.According to the state of WSN,the algorithm optimizes the dynamic communication network through continuous training,reasonably makes forwarding rules of routing decision and generates flow table information.Finally,we simulate the routing algorithm and verify the performance of routing decision in different data load and link environment.We compare our algorithm with traditional wireless sensor network routing algorithm.Experiments show that deep learning has application potential in some potentialities for large-scale network state space.Through the design of the scheme effectively reduce the energy consumption and improve the quality of data service.