Research on Robust Military Course of Action Generating and Optimization Methods
|School||National University of Defense Science and Technology|
|Course||Management Science and Engineering|
|Keywords||Factor state Environment state Course of action(COA) Action plan Resource matching plan Robustness of COA Cost of COA Complexity|
High uncertainty is one of the most important characteristics of modern warfare, itdemands that command and decision-makers must consider the uncertain factors thatmay affect the executing process or executing results when making a military course ofaction(COA). Modern warfare is high-tech warfare, the application of high-techequipment makes works such as intelligence investigation and information collectingbecame more convenient and efficient, however, these also bring quite a lot of dirty data.Command and decision-makers usually need to filter these dirty data in order to find outuseful information, this not only makes command and decision-makers tired out but alsonot conducive to make quick response to constantly changing situation. Therefore, thedifficulties of making modern military COA are dealing with huge amounts ofinformation and uncertainty of battlefield environment. Using computer to assistcommand and decision-makers to make military COA is one of the most importantmeans that settle above problems. Existing accurate algorithms for making militaryCOA usually cost too much time, intelligent algorithms and heuristic algorithms usuallycan not promise to find a COA that satisfies all the given constraints. In order to resolveabove problems, this paper focuses on investigating an efficient method for generatingmilitary COA in uncertainty battlefield environment.Concretely speaking, the main works of the paper are as following:1. Gives the process of generating military COA.The paper gives the process of generating military COA, which is good fordesigning efficient generating method. The paper defines several basic concepts thatrelated to the problem of generating COA in uncertain environment, including elementstate that denotes a single fact in objective world, environment state which isrepresented by a set of element states, military action which borrows the definition fromAI area, executing resource which represents military operational units, opposed actionwhich is executed by the other side in a warefare, and the military COA on which thepaper focus. The paper further analyzes the uncertainty in battlefield environment, anddefines the robustness of COA. Based on the above works, formally defines the problemof generating COA in uncertain environment. In order to investigate the problem moreefficient, the paper designs a process of generating COA, the robustness measurementof COA, the generating and optimization of action plan, and the generating andoptimization of resource matching plan.2. Designs a measurement method for COA’s robustness.The paper proposes a possibility achivement based method to measures therobustness of a military COA. The core of the method is an inference mechanism.According to the inference mechanism, an inference result can be obtained, based onwhich the robust value of the COA can be calculated. According to the different evaluating ranges, the paper divides the robustness of COA into two classes, one is theglobal robustness, and the other is the local robustness. The global robustness is used tocompare multiple COAs, and select a more suitable one according to the comparingresult. The local robustness is used to analyze the uncertain factors of COA, and theanalysis results can be used to improve the robustness of COA. In order to do theresearch more efficent, the paper further indicates a way to improve the robustness ofCOA, and gives the calculating method for the optimization of COA robustness.3. Designs the generating and optimization method of action plan based onrobustness.The paper designs a method for generating and optimizing an action plan. Thepaper first defines the concept of “Front Set”. After given theorems and correspondingproofs, the paper designs an initial action plan generating algorithm based on front set.Since the output of the PIPE algorithm is a sequence of actions, the paper furtherdesigns an algorithm to parallize a sequence action plan in order to improve theexecuting efficiency of a COA. The paper designs an algorithm to improve therobustness of a COA by adding redundant actions. The paper uses a multiple arms battlescenario to illustrate the process of each algorithm, and experimentally verifies theeffectiveness of the algorithms. The results of the experiments indicates that PIPEalgorithm is more advantage than general algorithms in time consuming, the degree ofoptimization and scalability etc., PARA algorithm can efficiently remove theunnecessary executing relation in an action plan, and REPA algorithm can efficientlyimprove the robustness through repairing an given COA.4. The generating and optimization of resource matching plan.The paper designs a method for generating and optimizing a resource matchingplan. The paper introdues the concept of resource matching conflict to denote thesituation that resource matching relations do not satisfy constrains. In addition, thepaper uses a directed graph to denote all possible confilicts in an action plan, anddesigns an algorithm(FindConf) for generating resource matching conflict graph. Afteranalyzing the complexity of the initial resource matching plan generating problem, thepaper designs a generating algorithm(CBRL). For a given action plan, CBRL algorithmmay add some additional executing order relation to ensure the best executing cost. Thepaper further proves that the complexity of the resource reparing problem is in NP,based on which it designs corresponding approximation algorithm(K-Greedy). The inputof the K-Greedy algorithm is a division of an action plan. The algorithm computes eachlocal resource matching plan corresponding to each division, and integrates all of theminto one as the output of the algorithm. K-Greedy algorithm has a good adjustability,and when the solution space is not null, the algorithm can promise to return a feasibleresource matching plan. The paper designs corresponding scenario to illustrate theprocess of FindConf algorithm, CBRL algorithm and K-Greedy algorithm, and gives experimental study. The experimental results show that all the algorithms can output anadvantage result, and each of them has good efficency and scalarbility.