【摘要】 针对非仿射高超声速飞行器控制问题,提出了一种基于强化学习的非线性反步控制方法。首先,基于高超声速飞行器的非仿射模型,设计了基于反步法的鲁棒跟踪控制器;其次,鉴于该控制器的性能和鲁棒性对参数比较敏感,借助深度确定性策略梯度的智能控制方法,对控制器进行了参数在线调整和控制指令补偿;最后,通过理论分析和仿真结果验证了当存在额外扰动和未建模动态时,所提出的智能控制器仍能保证攻角稳健跟踪期望目标。更多还原

【Abstract】 Aiming at the control problem of non-affine hypersonic vehicles,a reinforcement learning-based nonlinear backstepping control is proposed.Firstly,based on the non-affine model of hypersonic vehicles,a robust tracking controller based on the backstepping method was designed.Secondly,considering that the performance and robustness of the controller are sensitive to parameters,the parameters of the controller are adjusted online and control instructions are compensated by using the intelligent control method of deep deterministic policy gradient.Finally,theoretical analysis and simulation results show that the proposed intelligent controller can ensure the robust tracking of the desired signal with the presence of additional disturbances and unmodeled dynamics.更多还原