尾坐式无人机兼具传统固定翼飞机高速巡航与旋翼飞机垂直起降的优点，在军民领域均具有广阔的应用前景。其中，悬停转平飞和平飞转悬停过渡过程是该类无人机特有的飞行状态，但因俯仰角、飞行速度等的大范围变化，对其控制提出了极大挑战，是限制尾坐式无人机应用的关键。本文即针对过渡过程阶段飞行走廊构建、轨迹优化和鲁棒控制等典型问题展开研究，主要工作和创新点包括： （1）为解决传统配平飞行走廊未考虑尾坐式无人机过渡过程飞行速度及姿态动态变化的问题，设计了过渡过程动态飞行走廊。基于过渡过程受力/力矩机理，推导了能表征悬停转平飞“加速、低头”特性和平飞转悬停“减速、抬头”特性的数学模型，通过数值迭代建立了过渡过程动态飞行走廊；为防止过大的高度爬升，进而对走廊加入了爬升速度约束；给出了模型失配对动态飞行走廊的影响规律。 （2）为建立可兼顾安全裕度与飞行性能的过渡过程轨迹，提出了综合动态飞行走廊的多指标复合轨迹优化策略。为保证过渡过程轨迹与走廊边界的距离裕度，构造了飞行走廊的变权重参考线，建立了结合变权重参考线和性能需求的轨迹优化目标；在惩罚函数的基础上引入正弦变换，消除了轨迹优化问题的非线性约束，采用拟牛顿法和序列优化技术得到近似最优解；数值仿真表明，变权重参考线是安全裕度与飞行性能的很好权衡，所得过渡过程轨迹迎角更小，控制裕量更大。 （3）为保证模型失配时过渡过程轨迹的鲁棒性，建立了考虑耦合不确定性的鲁棒轨迹优化策略。考虑过渡过程初值及螺旋桨拉力系数和机翼气动系数不确定性，引入系统状态的期望和方差构造了鲁棒轨迹优化问题；基于Gram-Schmidt变换作不确定性解耦，采用多项式混沌展开作不确定性量化，将随机鲁棒轨迹优化问题转换为扩维的确定性轨迹优化问题； 基于Monte-Carlo测试的数值仿真表明，优化过程中不确定性的纳入显著提高了过渡过程轨迹的鲁棒性。 （4）为提升受扰时过渡过程轨迹跟踪的效果，提出了一种由轨迹优化控制量作前馈、LQR误差调节器作反馈，以及通过角加速度补偿姿态扰动的复合鲁棒控制方法。基于过渡过程长周期和短周期模态的动力学特性，给出了LQR控制权重矩阵参数整定方法，建立了闭环结构的指令调度策略；为获得准确的角加速度信号，构造了由标称力矩提供先验信息的角加速度估计模型，并采用Kalman滤波作角加速度估计；数值仿真和飞行实验验证了该复合控制方法的有效性和鲁棒性。
The tail-sitter unmanned aerial vehicle (UAV) combines the advantages of traditional fixed-wing UAVs for high-speed cruise ability and rotary-wing UAVs for vertical take-off and landing ability, which leads to broad application prospects in both military and civil fields. Among all flight stages of tail-sitter UAVs, the unique front and back transition processes are most challenging due to wide changes in the pitch angle, flight speed, and so on, which severely limits its application. Aim at this problem, this paper conducts research on the transition flight corridor construction, trajectory optimization, and robust controller design of tail-sitter UAVs. The main work and innovations are listed as follows: (1) To solve the problem that existing equilibrium transition corridors (ETC) do not consider the dynamic changes of the flight speed and attitude of the tail-sitter transition processes, a novel dynamic front transition corridor (DFTC) and a novel dynamic back transition corridor (DBTC) are developed. Based on the active forces/moments in the transition processes, mathematical models which can characterize the ``Accelerate and pitch down" features of the front transition and the ``Decelerate and pitch up" features of the back transition are developed. The DFTC and DBTC are then derived by numerical iteration. To avoid excessive large altitude climb during front and back transitions, a climb velocity constraint is further added. The effects of model mismatches on the transition corridors are also analyzed. (2) To establish a transition trajectory that can take into account both the safety margin and flight performance, a dynamic corridor-integrated multi-index transition trajectory optimization strategy is proposed. In order to ensure the distance margin between the transition trajectory and boundaries of the transition corridor, a weighted reference line of the dynamic transition corridor is developed. Based on it, a multi-index cost function is defined by combining the weighted reference lines and flight performance requirements. By introducing the sine transform on the basis of the penalty function, the nonlinear constraints of the trajectory optimization problem are eliminated. The approximate optimal solution is obtained by using the quasi-Newton method and sequence optimization technology. Numerical simulation shows that the weighted reference line is a good trade-off between the safety margin and flight performance. The resulting transitional trajectory has a smaller angle of attack and a larger control margin. (3) To ensure the robustness of the transition trajectory when the dynamic model is mismatched, a robust trajectory optimization problem considering coupling uncertainties is studied. By considering the transition initial state uncertainty as well as the propeller thrust coefficients and the wing aerodynamic coefficients uncertainties, the robust trajectory optimization problem is firstly constructed with the expectation and variance of stochastic system states. By employing the Gram-Schmidt transformation to decouple correlated uncertainties, and using polynomial chaos expansion (PCE) for uncertainty quantification (UQ), the stochastic robust transition trajectory optimization problem is expanded as a higher-dimensional deterministic transition trajectory optimization problem. Monte-Carlo simulation results show that the robustness of the derived transition trajectories is greatly improved. (4) To improve the trajectory tracking performance in the presence of uncertain disturbances, a composite robust transition trajectory tracking control law, which includes a trajectory optimization derived feedforward term, a closed-loop LQR feedback term, and an angular acceleration-based attitude disturbance compensation term, is designed. Based on the dynamic characteristics of the long- and short-period modes of the transition process, the weight matrix tuning rule of the LQR control is discussed. A closed-loop commands scheduling strategy is then established. In order to obtain accurate angular acceleration signals, an acceleration estimation model with prior information provided by the nominal moments is constructed. Based on the Kalman filtering, high-quality angular acceleration signals can be obtained. Numerical simulation and flight experiments demonstrate the effectiveness and robustness of the proposed control law.
