Research On ROHC Technology Based On Machine Learning

The bandwidth of the wireless link is the most precious and scarce resource in the wireless communication system,while a large amount of redundancy is contained in the network protocol header,therefore,the compression of the network protocol header can greatly save the bandwidth resources and improve the transmission efficiency of the wireless transmission.Robust Header Compression(Robust Header Compression,ROHC)algorithm is now recognized as ideal compression scheme on the wireless link.It use a variety of mechanisms to improve compression efficiency on the basis of guaranteeing the robustness of the algorithm,as a result,it has better compression performance on the wireless link which has high error rate,large channel attenuation,and high round-trip delay.However,the ROHC algorithm can only compress several protocol headers specified in the algorithm directly.To use the ROHC algorithm to compress other protocol headers requires artificial expansion of the ROHC algorithm,which will inevitably consume some manpower and time.So designing a more widely applicable header compression algorithm has great research significance.The main work of this paper is to design and realize a system that can automatically recognize the change characteristics of protocol fields,and compress protocol header according to the recognition characteristics with the combination of machine learning technology and ROHC algorithm.The main works of this paper are as follows: 1)ROHC algorithm and classification algorithm of machine learning are studied in depth;2)Lots of header information are collected as the training samples to set and test classification model and the decision tree algorithm and the Naive Bayes algorithm are used to establish the field changing characteristics classification model.Then the correct rate,average recall rate,average difference rate and the time taken to build the model are used as the evaluation indicators to compare and analyze the performance of the two models;3)Based on the classifier of field changing characteristics and ROHC algorithm thought,a robust header compression system based on machine learning was designed and realized.The system adopts the decision tree constructing classification model to classify each field changing characteristics of network protocol header and compresses protocol header according to network protocol field changing characteristics.1)A classification model of protocol header field changing characteristics based on decision tree is proposed,which can quickly and accurately classify the field changing characteristics of protocol headers;2)A header compression algorithm is proposed.This algorithm while ensuring the compression efficiency and robustness of the algorithm,it has wide applicability and can compress most of protocol headers;3)Designed and implemented a robust header compression system based on machine learning,which includes protocol description libraries,field changing characteristic classifier,compressor and decompressor modules.It has the function of automatically identifying the changing characteristics of the protocol field and compressing the protocol header according to the identification result.It can be used for header compression of most of network protocols.Finally,an test platform was constructed and an testl scheme is designed to test the system in this paper.Then,the test results are analyzed to show the effectiveness and superiority of the system in protocol header compression.