Steering control for underwater gliders

Steering control for an autonomous underwater glider （AUG） is very challenging due to its changing dynamic char- acteristics such as payload and shape. A good choice to solve this problem is online system identification via in-field trials to capture current dynamic characteristics for control law reconfiguration. Hence, an online polynomial estimator is designed to update the yaw dynamic model of the AUG, and an adaptive model predictive control （MPC） controller is used to calculate the optimal control command based on updated estimated parameters. The MPC controller uses a quadratic program （QP） to compute the optimal control command based on a user-defined cost function. The cost function has two terms, focusing on output reference tracking and move suppression of input, respectively. Move-suppression performance can, at some level, represent energy-saving performance of the MPC controller. Users can balance these two competitive control performances by tuning weights. We have compared the control performance using the second-order polynomial model to that using the filth-order polynomial model, and found that the tbrmer cannot capture the main characteristics of yaw dynamics and may result in vibration during the flight. Both processor-in-loop （PIL） simulations and in-lake tests are presented to validate our steering control performance.

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。