Research And Implementation Of Embedded Communication Middleware

The world-wide use of embedded systems makes embedded system interconnection network become an inevitable necessity.However,due to the heterogeneity and peer-to-peer characteristics of embedded networks, the traditional network protocols can not be utilized directly by the embedded system interconnection.So it becomes a pressing demand to build an embedded system interconnection platform which can shield the heterogeneity of embedded systems and make use of the resources in embedded systems equally,evenly and reasonably.In this thesis,firstly the existing standards and technologies of the interconnection platform in two embedded systems application environments:digital home network and industrial control network are analyzed.In view of the problems and deficiencies of these technologies and standards,such as disunity,poor compatibility,being unable to seamless connect and protection of intellectual property rights,the concept and framework of embedded communication middleware are proposed to resolve the communication problems in embedded system interconnection.The hardware-independent driver layer and OS interface layer are utilized in this embedded communication middleware to shield different hardware and operating systems at the bottom.While the internal callback functions are provided to give application interfaces for the upper software.Thereby,the middleware is cross-platform and can support a wide range of communication modes and applications.Based on the basic framework of embedded communication middleware,the routing algorithms and protocols in its routing layer are deeply studied in this thesis.In view of limited resources of embedded system and peer-to-peer characteristics of embedded networks,combined with Markov decision process,routing algorithms which are distributed and using local decision are adopted.Improved model and the corresponding algorithms of unicast,dynamic unicast,multicast,and QoS routing—MFHDR,DMFHDR,MMFHDR and MQoSR—based on Markov Finite Horizon Decision model,are constructed for embedded system interconnection to respectively solve the simple command transmission, multicast transmission,and QoS streaming multimedia transmission in embedded network.It is proved by the complexity analysis and simulation results that:the algorithms above have the feature of distributed parallel computing which can distribute the computation and storage requirement of routing to each device,so that it can balance the workload between all the devices without increasing the workload of a single one.And also because the algorithms all consider the multiple dynamic factors of embedded systems,they can compute more reliable and efficient routing paths,which decrease the drop packet ratio and busy time of links.Another key issue of embedded communication middleware—the queuing and scheduling of its data packets—have also been researched in this paper.A new method to give packets different priorities and an M/G/1 queue algorithm based on these priorities are put forward.The theory performance indexes are analyzed,and the queuing system procedure and strategy are provided.The experimental result shows that this queuing schedule algorithm without preemption based on multi-priorities can effectively guarantee the real-time character of packets with high priority and the fairness of packets with low priority and reduce the queue length.Meanwhile,it is simple to control and takes up small memory and buffer space.Finally,some researches and practice are done in implement level of embedded communication middleware in this thesis.Parts of codes and key functions are designed and compiled;a reduced multi-threads model is proposed and some problems during the application are discussed.