【作者】 胡中华；

【导师】 赵敏；

【作者基本信息】 南京航空航天大学 ， 测试计量技术及仪器， 2011， 博士

【摘要】 无人机航迹规划是根据任务目标规划出满足约束条件的飞行轨迹，是无人机任务规划系统的关键组成部分。本文得到航空科学基金项目（2009ZC52041）资助，对无人机静态航迹规划、突发威胁下航迹规划、无人机多机协同航迹规划及航迹平滑等关键技术进行了研究。针对无人机航迹规划问题所需满足的约束条件，建立了无人机机动性能约束模型和威胁约束模型，其中前者包括最大航程、最大爬升角、最小转弯半径及最小步长；后者主要包括雷达、导弹、高炮、大气和地形等威胁模型。此外，考虑到无人机可以利用地形进行规避风险，因此将高度作为航迹综合代价之一。标准蚁群算法应用到无人机航迹规划问题中，状态转移策略仅根据信息素及启发因子按概率选择，存在盲目选择且难以快速找到目标节点的问题。本文在标准蚁群算法的状态转移策略中引入导引因子，并通过设定最大航迹节点数，解决了标准蚁群算法难以找到目标节点的问题。此外，在蚁群算法中引入随机蚂蚁子群，可以扩大搜索空间并增加解的多样性，使得算法可以获得更精确的解。仿真结果表明：改进的算法较原算法具有一定的优势。无人机实际飞行中如果存在突发威胁分布的情况，必须进行航迹重规划，以便规避威胁。为满足作战时效性，要求重规划所采用的算法必须具有实时、高效的特点。本文根据蜂群算法邻域搜索的特性，以参考航迹的突发威胁段作为引领蜂航迹，跟随蜂仅在参考航迹的突发威胁段进行邻域搜索，而不需要对整条航迹搜索，由此可以快速获得修正航迹段，并替换原突发威胁航迹段。整个飞行过程中，无人机根据获得的威胁信息，不断修正参考航迹，直至达到目标节点。仿真结果表明：该算法在局部航迹修正方面较蚁群算法有一定的优势。针对无人机协同航迹规划问题提出两级规划算法。该方法将协同航迹规划划分为航迹规划层和协同规划层。其中，航迹规划层首先设置满足每架无人机攻击角度的攻击节点集，然后采用智能优化算法得到各候选节点的相应的候选最优航迹集；协同规划层通过设计不同无人机的候选航迹之间的协同变量和协同函数，确定各候选综合协同代价最小的方案。最后，分别对协同会聚攻击及协同轮流攻击两种攻击策略进行仿真，结果表明：该方法可以获得全局最优备选航迹方案集，并能准确生成符合时空协同要求的航迹。无人机初始航迹规划仅考虑了部分无人机机动性能约束，所获得的航迹仅能满足战术运筹，难以满足飞行性能约束，因此必须进行航迹平滑。本文根据航迹平滑问题的参数分布不均匀的特点，为保证平滑后的航迹代价无显著变化，提出了基于三次非均匀B样条曲线插补的航迹平滑方法。仿真结果表明：采用本文算法平滑后的航迹转弯半径均大于无人机所允许的最小转弯半径，整体过渡自然，航向没有突变。该平滑航迹不但经过全部航迹节点，而且逼近原航迹曲线，因此航迹的综合代价较原航迹无显著变化。最后，针对无人机航迹各代价权重在起初往往取平均值或根据经验来确定，存在主观性较大的缺陷，本文充分考虑各指标之间的关系，引入离差最大化法和信息熵法来求解权重。此外，在多个方案进行优选决策时，进一步考虑各因素之间相互关联的系统特性，引入灰色关联分析理论，构建无人机航迹规划方案的优选模型。最后，将此优选模型用于实际航迹规划问题，从航迹方案集中挑选出综合性能最优的航迹，避免了传统凭经验选型的主观性和随机性。更多还原

【Abstract】 UAV path planning is to plan the flight path and meet the constraints. It is based on the missionand is a key component of UAV mission planning system. This paper is supported by the AviationScience Foundation (2009ZC52041). This paper studies mainly some key issues，include the UAVPath Planning, including UAV static path planning，path planning with pop-up threats，cooperativepath planning for UAVs and path smoothing.To meet the constraints of path planning problem, the paper establishes the UAV dynamicconstraint models and the threat constraint models. The former includes the maximum range, themaximum angle of climb, the minimum turning radius and the smallest step. The latter includesterrain threat, radar threat, missile threats, artillery threat, and atmospheric threat. In addition, takingthe terrain into account to avoid the risk of UAV, so this paper regards height as one of the cost.When the standard ACO algorithm is applied to UAV path planning problem, thetransition-strategy of state makes a choice by the probability based only on inspiration factor andpheromone. It is easy to do a blind selection and it is difficult to quickly reach the target node. In thealgorithm of this paper, the guidance factor is introduced in the state transition strategy. By setting themaximum number of path, it solves the problem that the number of nodes is not fixed and it isdifficult to find the target nodes. Furthermore, random ant subgroup is introduced into the algorithmand it can expand the search space and increase the diversity of solutions to obtain more accuratesolution. The simulation results show that the performance of the improved algorithm is better thanthe original algorithm.In addition, the pop-up threat may appear in the flight process of UAV, it’s necessary to generatethe new path quickly to avoid the threat. In order to meet the time limit, the algorithm must bereal-time, high efficiency. According to the characteristics of neighborhood search of Artificial BeeColony algorithm, path segment with sudden threats is considered as the lead path. Neighborhoodsearch is just done at the reference path section with sudden threats by following bee. It will not bedone at other reference path section. Therefore, we can obtain the optimal trajectory segment quickly,and replace the original path segment with Pop-up threat with a new reference trajectory. In thethroughout the flight, UAV determine path segments with the Pop-up threats based on threatinformation and modify the reference trajectory repeatedly until reaches the target node. Simulationresults show that: This Algorithm has more advantages in terms of local path modification than ACO algorithm.To solve the problem of cooperative path planning for UAVs, two-stage planning algorithm isproposed. In this method, cooperative path planning is divided into route planning layer andcooperative planning layer. Among them, the path planning layer, candidate attack node is determinedby setting the attack angle of each UAV. The corresponding candidate optimal path set of is attainedby intelligent optimization algorithm. In cooperative planning layer, coordination functions andcoordination variables of each candidate path are designed. The cooperative solution with minimumcost is determined. Finally, two attack strategies，cooperative converge attacks and cooperative tookturns attacking， are researched separately. Simulation results show that: the global optimumalternation can be attained by the method and it can generate accurate path which meet requirementsof space-time collaboration.UAV mobility constraints are not all taken into accounted during the Initial path planning, so theinitial path can only meet the tactical operations. But for the actual flight, it is often difficult to meetthe maneuverability performance constraints. Therefore, it is necessary to smooth the path and threenon-uniform B-spline curve interpolations are proposed to smooth the path. It will also ensure thatthere are no significant change between the old path and the smooth one. Two-dimensional andthree-dimensional initial paths are simulated. The results show that turning radius of the smoothedpath is greater than the minimum turning radius of the UAV. And the overall transition is natural;heading of the path has no mutations. The smooth path not only passes all the path nodes，but alsoapproaches the original path curve. Therefore, its comprehensive cost has no significant change bycontrasted with the original path.Finally, since weight of indicator is often determined based on average and experience，it issubjective. This paper fully considers the relationships between each index, weights of them aredetermined according to objective method. Deviation maximization method and information entropyare introduced into solve weights. In addition, in optimizing decision-making of multiple solutions,interrelated system characteristics between various factors are further considered. Optimizationdecision-making system with multi-objective and multiple constraints is established. Gray relationalanalysis (GRA) is introduced to construct the optimization model of path planning programs. It isused to path planning problem to avoid the subjectivity and randomness of traditional method byexperience.更多还原