This article presents an architecture for estimation of a flow field using a hypothetical oceanographic vehicle that is guided along paths of high flow-field observability, a concept quantifying the informativeness of a path. Sampling trajectories that pass close to saddle points along separating boundaries of invariant sets provide high observability of flow-field parameters. The estimation and control framework consists of a model predictive controller that utilizes a measure known as the empirical augmented unobservability index to select from candidate trajectories generated by steering the vehicle to separating boundaries of invariant sets. Empirical augmented observability extends empirical observability to account for prior uncertainty when performing path planning based on observability. While following a selected trajectory, the vehicle takes measurements of its position (e.g., GPS measurements) and accounts for its own actuation to produce Lagrangian measurements. The vehicle assimilates this Lagrangian data in a Gaussian mixture Kalman filter, which is a nonlinear/non-Gaussian filter, to recursively improve its map of the flow field. Using the posterior uncertainty of the map, the vehicle plans new candidate routes and continues to sample adaptively. The performance of this estimation architecture is demonstrated for a simplified dynamic model of a pair of ocean eddies.

中文翻译：

---

由驱动拉格朗日传感器对潜在流的非高斯估计，通过增强的可观察性转向分离边界

本文介绍了一种使用假设的海洋学车辆来估计流场的架构，该车辆沿着高流场可观测性的路径引导，这是一个量化路径信息量的概念。沿着不变集的分离边界通过靠近鞍点的采样轨迹提供了流场参数的高度可观察性。估计和控制框架由模型预测控制器组成，该控制器利用称为经验增强不可观察性指数的度量从通过将车辆转向分离不变集边界而生成的候选轨迹中进行选择。经验增强可观察性扩展了经验可观察性，以在基于可观察性执行路径规划时考虑先验不确定性。在遵循选定的轨迹时，车辆测量其位置（例如，GPS 测量值）并考虑其自身的驱动以产生拉格朗日测量值。车辆在高斯混合卡尔曼滤波器（非线性/非高斯滤波器）中吸收该拉格朗日数据，以递归地改进其流场图。利用地图的后验不确定性，车辆规划新的候选路线并继续自适应采样。这种估计架构的性能在一对海洋涡流的简化动态模型中得到了证明。利用地图的后验不确定性，车辆规划新的候选路线并继续自适应采样。这种估计架构的性能在一对海洋涡流的简化动态模型中得到了证明。利用地图的后验不确定性，车辆规划新的候选路线并继续自适应采样。这种估计架构的性能在一对海洋涡流的简化动态模型中得到了证明。