Research On Uncertain Complex Event Processing In Agriculture Cyber-Physical Systems

The Cyber-Physical System(CPS)is the next generation of intelligent system,which can realize the close integration and coordination of computing resources and physical resources by the deep integration of computation,communication and control technology.CPS technology has been used in medical,energy,transportation and other important areas of development,and has broad application prospects.The organic combination of CPS technology and agriculture is called Agriculture CPS.Agriculture CPS may contain a wide variety of interconnected devices,which continue to produce huge amounts of original data.At the same time,most of the current CPS systems are distributed and the data from different levels of data sources are massive,heterogeneous and disperse.However,the traditional data processing technology is difficult to deal with them effectively,so event based processing method can be accepted.In the event stream of Agriculture CPS collection,the meaning of a single event is limited,and the users are more concerned about the sequence of events conforming to a specific pattern that reflect the logic of agricultural intelligent control.How to identify the sequence of events from the event stream is a difficult problem in Agriculture CPS data processing.Complex event processing(CEP)technology as the core technology of CPS can solve this problem effectively.In addition,noise,sensor errors,clock synchronization,network load,delay and other reasons may cause the uncertainty of agricultural data.Therefore,it is necessary to study the uncertain complex event processing technology.The uncertainty of events can be quantified as probabilities,so the processing of uncertain event flow is transformed into the processing of probabilistic event flow.Probabilistic complex events are generated by the combination of multiple probabilistic atomic events in a specific pattern.In the process of combination,how to calculate the probability of complex events from the generation of the atomic event probability and how to match the events in a specific mode are the two main problems that need to be studied.In this paper,based on the in-depth analysis of the current research status and challenge of CEP technology in CPS,to solve these two problems,several aspects have been studied deeply using Agriculture CPS as the background and the greenhouse as examples,which contain the probability calculation of complex events,tree based complex event detection algorithm with uncertain attributes,and probability time Petri net based complex event detection model with uncertain time.The main work of this paper includes the following aspects.(1)Researches on the probability calculation of complex events in Agriculture CPS are performed,and a method of probability calculation based on approximate provenance is proposed.Uncertain complex event processing is to detect the sequence of events that satisfy the probability threshold,so how to calculate the probability of complex events composed of the flow of uncertain atomic events is one of the main challenges in processing complex events.Data provenance management theory is added in this paper.In order to cope with the continuous generation of massive probabilistic atomic events,sufficient approximate provenance algorithm is introduced and discrete polynomial approximate provenance algorithm is proposed.The analysis and comparison of the two algorithms are carried out.Experimental results show that two algorithms can calculate on the compressed data and reduce the amount of computation to improve calculation efficiency.Specially,the latter is better,which compression ratio is higher than the former.(2)The problem of tree based uncertain complex event detection algorithm in Agriculture CPS is studied.The PUCEP algorithm with uncertain general attribute and the coded ESI-tree solution with uncertain time attribute are proposed.Sensor error,acquisition accuracy,network communication technology and so on in Agriculture CPS may cause uncertainty of events' attributes.PUCEP algorithm for the complex event detection of event streams with uncertain general attributes is proposed,which combines the the binary tree theory and NFA algorithm.Then the probability threshold is introduced to optimize it.Experimental results show that the optimized algorithm can gain significant improvement in performance.Time is a special attribute of an event.In view of its particularity,a tree based solution called ESI-tree algorithm is proposed,and the tree is compressed according to its characteristics,and then the coding method is used to optimize it.The experimental results show that the coded detecting algorithm has the advantages of efficiency and memory usage.(3)The problem of Petri net based uncertain complex event detection modeling in Agriculture CPS is studied,and a method for complex event detection model based on probabilistic time Petri net is proposed.Complex events in Agriculture CPS are formed by the combination of atomic events,and an atomic event is closely related to the collection,transmission,conversion process of various parts in CPS.In this process,atomic events' time uncertainty is generated due to system error and random error.Aiming at the uncertainty of time,using a greenhouse as a case,probabilistic time Petri net is used to model atomic events of the system and form atomic events model module,then probabilistic time Petri net modules of atomic events are combined to form complex event model.Using this model,the semantics of complex events can be accurately analyzed and misjudgment and leakage judgment of complex events in the matching process can be detected.Finally,the correctness of this method is proved by case analysis and it is proved that the higher the time delay probability is,the greater the misjudgment possibility of complex event detection is,and the higher the system cost is.