多智能體的戰(zhàn)術(shù)行為決策研究及應(yīng)用.pdf

1、南京航空航天大學碩士學位論文多智能體的戰(zhàn)術(shù)行為決策研究及應(yīng)用姓名：宋琛申請學位級別：碩士專業(yè)：計算機科學與技術(shù)指導(dǎo)教師：燕雪峰2011-03多智能體的戰(zhàn)術(shù)行為決策研究及應(yīng)用 II Abstract Tactical behavior is important to military confrontation, also a problem of agent collaboration. Multi-agent task’s core

2、is coordination and implementation. Complex task is difficult to implement for a single agent, requiring groups together to complete coordination and cooperation with each other. How to organize the body posed by a numbe

3、r of agents, and how groups such agents to achieve coordination and cooperation in the confrontation is an important theoretical and practical significance of the research topic. The work is from research cooperation of

4、China State Shipbuilding Corporation and Nanjing University of Aeronautics and Astronautics, which through the confrontation of agent to research the battlefield tactics of naval warfare. The study has found that the deg

5、ree of combat confrontation can influence the simulation results. The study research tactical coordination behavior through the agent’s acts, first study cooperative navigation algorithm of the formation, then study the

6、offense and defense decision-making when formation moving. Finally, we prove the feasibility and effectiveness of study by simulation experiments. The theory of potential field and particle swarm Optimization is used to

7、study formation problem and path planning problem. Through a new joint control algorithm we can get better formation control. Traditional PSO for Path planning can not adjust the number of nodes according to in different

8、 environments, so that the search efficiency is low, even in some terrain are not feasible solutions. This paper proposes a particle swarm optimization algorithms based on variable-dimensional for path planning. The algo

9、rithm dynamically changes the dimensions of particles by controling the number of nodes, to accelerate the convergence speed. In formation control, the design of formation evaluation function determines the state of form

10、ation, then using different control strategies. The algorithm enhances capabilities to dynamicly avoid the threat of obstacles by different control strategies. The the artificial potential’s deficiencies is that the meth

11、od is easy to fall into local extreme, so the paper propose a virtual potential point algorithm to avoid obstacle and restructure the formation. Simulation results show the effectiveness of the algorithm. In order to imp

12、rove the effectiveness of cooperative attack and defense, the paper researches immune algorithm in the target allocation problem. The current target allocation algorithm is more for once fire coverage, so the combat troo