论文部分内容阅读
水下滑翔机动力学特性随有效载荷及外形变化而变化,其航向控制富有挑战性。解决方法是使用在线系统辨识算法捕捉当前动力学特性,更新运动模型。为此,我们设计了一个在线多项式辨识器,不断更新当前动力学模型,同时用一个自适应模型预测控制器计算并输出最优化的控制指令。该控制器根据用户自定义的指标函数,使用二次规划方法得到最优控制指令。该指标函数由两项组成,一项用来表达轨迹跟踪性能,一项用来表达输入指令抑制性能。输入指令抑制性能一定程度上可以代表该控制器的能量消耗性能。设计师可以通过调节这两项的权重,平衡两个控制器的性能。比较二次与五次多项式模型的控制效果,发现:二次多项式模型不足以表达无人机的动力学特性,且控制结果易发生剧烈波动。硬件在环模拟以及湖试结果验证了控制器性能。
Underwater glider dynamic characteristics with the payload and shape changes, the heading control is challenging. The solution is to use the online system identification algorithm to capture the current dynamics and update the motion model. To this end, we designed an on-line polynomial recognizer that continuously updates the current dynamics model and uses an adaptive model predictive controller to calculate and output the optimal control commands. The controller uses the quadratic programming method to get the optimal control instruction according to the user-defined index function. The indicator function consists of two components, one for tracking performance and the other for expressing input command rejection performance. The input command suppression performance can, to some extent, represent the controller’s energy consumption performance. The designer can balance the performance of the two controllers by adjusting the weight of both. Comparing the control effects of quadratic and quintic polynomial models, we find that the quadratic polynomial model is not enough to express the dynamics of UAV, and the control results tend to fluctuate violently. Hardware in-loop simulation and lake test results verify controller performance.