Sibley School of Mechanical and Aerospace Engineering, Cornell University, Ithaca, NY 14850, USA.
Sensors (Basel). 2021 Jul 5;21(13):4603. doi: 10.3390/s21134603.
To aid the development of future unmanned naval vessels, this manuscript investigates algorithm options for combining physical (noisy) sensors and computational models to provide additional information about system states, inputs, and parameters emphasizing deterministic options rather than stochastic ones. The computational model is formulated using Pontryagin's treatment of Hamiltonian systems resulting in optimal and near-optimal results dependent upon the algorithm option chosen. Feedback is proposed to re-initialize the initial values of a reformulated two-point boundary value problem rather than using state feedback to form errors that are corrected by tuned estimators. Four algorithm options are proposed with two optional branches, and all of these are compared to three manifestations of classical estimation methods including linear-quadratic optimal. Over ten-thousand simulations were run to evaluate each proposed method's vulnerability to variations in plant parameters amidst typically noisy state and rate sensors. The proposed methods achieved 69-72% improved state estimation, 29-33% improved rate improvement, while simultaneously achieving mathematically minimal costs of utilization in guidance, navigation, and control decision criteria. The next stage of research is indicated throughout the manuscript: investigation of the proposed methods' efficacy amidst unknown wave disturbances.
为了辅助未来无人舰艇的开发，本文研究了将物理（嘈杂）传感器和计算模型相结合的算法选项，以提供有关系统状态、输入和参数的附加信息，强调确定性选项而非随机选项。使用庞特里亚金（Pontryagin）对哈密顿系统的处理方法来构建计算模型，从而根据所选算法选项获得最优和近最优结果。反馈用于重新初始化两点边值问题的初始值，而不是使用状态反馈来形成由调谐估计器校正的误差。提出了四种算法选项，并带有两个可选分支，将所有这些与包括线性二次最优在内的三种经典估计方法进行了比较。运行了超过一万次模拟来评估每种拟议方法在典型嘈杂状态和速率传感器中对植物参数变化的脆弱性。所提出的方法在状态估计方面提高了 69-72%，在速率改进方面提高了 29-33%，同时在制导、导航和控制决策标准中实现了数学上最小的利用成本。整篇论文都指出了下一阶段的研究方向：研究未知波浪干扰下提出的方法的效果。
