高超声速飞行器滑模控制参数整定方法设计

本文针对高超声速飞行器滑模控制人工试凑的参数整定方法较为繁琐、效率低的问题，改进了连续动作学习自动机（Continuous action reinforcement learning automata, CARLA）算法并将其应用于滑模控制参数整定问题中，改进后的算法通过对回报函数的设计有效克服了常规CARLA算法收敛速度慢、易受干扰、求解效率低的问题。该算法引入控制性能指标评价函数，在迭代中学习阶跃响应的经验数据，实现了控制参数自整定。仿真表明，对于阶跃响应问题，本文提出的算法能够在100次迭代中整定出一组高品质的控制参数，完成对给定指令的快速准确跟踪，与遗传算法、模拟退火算法相比，在求解速度上具有显著优势。由于该方法不依赖于模型，除了滑模控制器参数整定外对其他控制方法的控制参数整定问题也有一定适用性，具有推广应用价值。

Design on Parameters Tuning Method of Sliding Mode Control on Hypersonic Aircraft

In this paper, an algorithm is proposed to solve the problem of tedious and inefficient tuning of sliding mode control parameters for hypersonic aircraft. This algorithm improves the continuous action reinforcement learning automata(CARLA)algorithm by redesigning the return function, and overcomes the original algorithm problems of slow convergence, being easy to receive disturb and low efficiency. The algorithm introduces the evaluation function of control performance index, learns the empirical data of step response in iteration, and realizes the self-tuning of control parameters. The simulation results show that for the step response problem, the algorithm proposed in this paper can get a set of high-quality control parameters in 100 iterations to track the given instructions quickly and accurately. Compared with the genetic algorithm and simulated annealing algorithm, it has significant advantages in solving speed. This method does not depend on the model, and besides the sliding mode controller, it is also applicable to the control parameters tuning of other control methods. In general, it has the value of promoting applications.