【作者】 张胜祥；

【导师】 裴海龙；

【作者基本信息】 华南理工大学 ， 控制理论与控制工程， 2009， 博士

【摘要】 本文研究了小型无人直升机的航迹规划，建立了基于滚动时域控制和混合整数线性规划（RHC-MILP）的航迹优化算法。这种算法特别适用于环境事先未知，需要在线逐渐探测的情形。本文利用数学规划建模语言AMPL以及采用高性能的商业优化求解软件CPLEX进行计算机仿真验证。仿真结果表明，对于一个复杂环境下的航迹规划问题，基于RHC-MILP的航迹优化算法能够实时求解出满足飞行器动态的最优航迹。本文从以下几个方面进行了研究。首先，本文研究了飞行器在城市环境中飞行的航迹规划。讨论了三维建筑物作为障碍物的障碍物回避。建立了基于RHC-MILP的航迹规划算法。通过引入逻辑变量和连续变量的混合形式的线性约束来描述障碍物回避约束，对飞行器的动态特性进行线性近似，以最小时间和其它性能指标作为代价函数，建立混合整数线性规划，并采用滚动时域控制策略求解，仿真结果显示此算法能够实时规划最优航迹。第二，本文研究了飞行器在山地环境中飞行的航迹规划。讨论了飞行器实现地形回避和地形跟随的方法。建立了基于RHC-MILP的航迹规划算法。本文提出一个新的方法—结合不规则三角网（TIN）和MILP描述地形回避。通过在代价函数中增加一项高度代价，选取适当的权因子，实现地形跟随。采用滚动时域控制策略，以及在地形回避约束中只考虑优化时间窗口范围内的局部地形，极大地减少求解时间。基于随机地形的仿真验证了算法的实时性和有效性。第三，本文研究了直升机三维机动飞行的航迹规划。把飞行器在悬停，前飞等飞行模态之间的切换和各机动动作建模为混合自动机，把航迹最优化问题看作一个序列决策过程。用连续决策变量实现各飞行模态的连续优化，模态选择、模态切换和机动动作的触发通过逻辑决策变量来实现。此决策问题可以采用基于混合整数线性规划的优化算法解决。最后本文研究了多飞行器的协调飞行航迹规划。建立了基于DRHC-MILP的航迹规划算法。采用分布式滚动时域控制策略，把多飞行器组成的飞行编队的航迹规划问题，分解成多个单一飞行器的航迹优化子问题。飞行编队中的每一个飞行器在线求解一个小规模的优化子问题而规划其自身的飞行轨迹，各单个飞行器作出的规划轨迹除了满足地形回避，还满足碰撞回避。各优化子问题通过分组可以并行计算。更多还原

【Abstract】 This thesis studies the trajectory planning for the small-scale unmanned helicopters andpresents several trajectory optimization algorithms based on Receding Horizon Control andMixed Integer Linear Programming(RHC-MILP). The algorithms are specially suitable for thecase that environment is unknown ahead of time and needs to explore online. The algorithms aretested and demonstrated by computer simulation with the mathematical programming languageAMPL and the powerful commercial optimization solver CPLEX. The simulation results showthat, for the trajectory planning of the vehicles in complicated environment, the algorithms areapplicable of computing optimal trajectories in real-time. The research fields in this thesis areas follows.Firstly, the thesis studies the trajectory planning for the vehicles ?ying in the city. Theobstacle-avoidance is discussed which the obstacles is 3-dimension buildings. The algorithmof trajectory planning based on RHC-MILP is presented. The obstacle-avoidance constraintscan be formulated as linear constraints by introducing logic variables combining continuousvariables. The dynamics of the vehicles is linearly approximated. The cost function is aboutminimal time or/and other performance index. Then trajectory optimization problem can be for-mulated as MILP forms. By solving the MILP and using receding horizon strategy, an optimaltrajectory from start to goal is planned online. The simulation result shows that the algorithmcan produce optimal trajectory in real time.Secondly, the thesis studies the trajectory planning for the vehicles ?ying in the moun-tainous region. The terrain-avoidance and terrain-following of vehicles are discussed. Thealgorithm of the trajectory planning based on RHC-MILP is presented. A novel method is usedthat combines triangulated irregular network(TIN) and MILP describing terrain-avoidance. Thecost function has an item about the altitude cost, a suitable weight could keep the trajectory ofthe vehicle follows the terrain. By taking receding horizon strategy, and only considering thelocal terrain within the planning horizon, the computing time is greatly reduced. The simulationbased on random terrain demonstrates the the real-timeness and validness of the algorithm.Thirdly, the thesis studies the trajectory planning of agile vehicles. The transition amonghover, cruise and other ?ight modes as well as maneuvers are modeled as an automata. Tra-jectory planning is viewed as a sequential decision process. The continuous decision variablesare associated with the optimization of the ?ight modes, while logic decision variables are as-sociated with mode-choosing, mode-transition and maneuver execution. The guidance decisionproblem can be formulated as the MILP form. Finally, this thesis studies the trajectory planning for multi-vehicle’s cooperative ?ights.The algorithm of the trajectory planning based on DRHC-MILP is presented. The distributedalgorithm of the trajectory optimization for multi-vehicle ?eet breaks the optimization intosmaller subproblems. Each vehicle in the ?eet plans its trajectory by solving a reduced sizeoptimization problem online. Each trajectory planed by the single vehicle satisfies all con-strains of collision avoidances as well as terrain avoidances. The optimization subproblems canbe solved within a group in parallel.更多还原