Font Size: a A A

Research On Tactical Intention Recognition Based On Rule Discovery And Bayesian Reasoning

Posted on:2016-09-06Degree:DoctorType:Dissertation
Country:ChinaCandidate:S GeFull Text:PDF
GTID:1316330542974117Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Battlefield situation varies from minute to minute during the military warfare and opportunities for combat are fleeting.The quick acquisition,correct understanding and deduction of valuable data from the massive battlefield situation information for accurate recognition of tactical intention of enemy target is a kind of important basis for battlefield command and decision and also a key factor for the result of war.At present,with the high speed development of technology,the informatization and intelligentialization of military operations become obvious.The studies and application of battlefield information fusion technology also gradually shift to high level.The battlefield situation analyzing and evaluating technology which takes the inference and recognition of tactical intention as representative are becoming a research hotspot.Taking national defense beforehand research projects during "the Eleventh Five-Year Plan" and "the Twelfth Five-Year Plan" as the background,based on analyzing the relevant concepts and actualizing processes of combat mission and tactical intention,this thesis proposes SBN(Series Bayesian Network)model for expressing and reasoning of planning and decomposing processes,and establishes DSBN(Dynamic Series Bayesian Network)model for tactical intention recognition based on SBN model;then,according to the constructing problems of DSBN model,this thesis extents expression ability of MEBN(Multi-Entity Bayesian Network)model,and proposes the constructing algorithm of DSBN model based on extended MEBN;aiming at the acquisition of extended MEBN used for DSBN construction,the thesis devises the discovery processes of inference knowledge for tactical intention recognition with theories of statistics and data mining innovatively.The main innovative works of this thesis are as follows:(1)SBN(Series Bayesian Network)model to express processes of plan carrying out is proposed and DSBN(Dynamic Series Bayesian Network)model for expressing and reasoning of tactical intention recognition is established.The layout process which is the important phase in carrying out of plan is analyzed.And the SBN model to express this phase is proposed.Correspondingly,the relationship described by this model between some states of parent random events and some series of probability states of child events cannot be expressed by classical BN model,DBN model and other similar models.SBN is an appropriate model for process of planning,and is a necessary structure for expressing and reasoning tactical intention recognition.At the same time,the probability computing method of SBN model based on the computation of event probability state sequence with child events transform rules is researched and given out.The planning and carrying out processes of combat mission and tactical intention are analyzed.These two phases are expressed with SBN and DBN model respectively,and DSBN model is established which can be widely used for planning and solving processes.Inference of DSBN model is presented and is compared with BN and DBN model qualitatively;the model validation shows that with the expressing and reasoning ability of planning phase of SBN model,DSBN are effective models for planning recognition problems such as tactical intention recognition.(2)The MEBN(Multi-Entity Bayesian Network)model for the need of SBN construction is extended and a kind of construction algorithm of DSBN model based on EMEBN model is proposed.The expression ability of the probability inference knowledge of MEBN model and the construction process of BN model are introduced;the deficiencies of MEBN's expression for layout process are analyzed and then MEBN is extended.The expression ability of transition processes and planning processes of random events probability state to MEBN is enhanced,the EMEBN(Extended MEBN)model is established,with which DSBN can be expressed and constructed.The constructing algorithm of DSBN model is proposed based on the inference knowledge of EMEBN model and the given background knowledge of battlefield situation and the time and space complexity of the algorithm are analyzed.Qualitative comparison between the constructing algorithm based on EMEBN and the classical learning algorithm are made which shows that the classical learning algorithm does not have the ability of building SBN structure,while the algorithm based on EMEBN can preferably construct the DSBN model based on fragments of inference knowledge about enemy target used for tactical intention recognition.It demonstrates and shows that the flexibility of the algorithm which can add,delete and modify the knowledge fragments as needed easily and also can construct different DSBN model according to different battle field situations.(3)The discovery processes of inference knowledge for tactical intention recognition is designed by using theories of statistics and data mining.Aiming at the problem that the inference knowledge of enemy target to constructing DSBN model for tactical intention recognition cannot be obtained directly,this thesis creatively uses statistics and data mining theories to find inference knowledge used for constructing DSBN model and tactical intention recognition,proposes a kind of difference measurement of multi-dimension vector,and applies clustering method,membership grade method,transform frequency statistics and sequential pattern mining algorithm for discovery inference knowledge of relationship between random events of battlefield situation,timing state transform rules of random events and planning sequential pattern rules of random events,in order to meet the needs of forming EMEBN model from these rules and in order to construct DSBN model used for tactical intention recognition.Experiments and simulations show that the way to gain tactical intention recognition with DSBN model constructed from inference rules which are extracted from sample or historical data is feasible and available.Models and algorithms concerned can be the foundation and support for related researches and applications.
Keywords/Search Tags:SBN, DSBN, tactical intention recognition, EMEBN, knowledge discovery
PDF Full Text Request
Related items