20260625:很多新内容

This commit is contained in:
2026-06-25 14:08:47 +08:00
parent 91fac5b6fc
commit 6021dea160
375 changed files with 19263 additions and 251 deletions

View File

@@ -0,0 +1,30 @@
---
title: "扩展 Kalman 滤波"
created: 2026-06-22
updated: 2026-06-22
type: concept
tags: [state-estimation, filtering, nonlinear-systems]
sources: [nano-filter]
---
# 扩展 Kalman 滤波
Extended Kalman Filter (EKF) 是最早的非线性 [[kalman-filter|KF]] 扩展,通过对非线性函数做一阶 Taylor 展开实现局部线性化。
## 核心机制
对非线性系统 $x_t = f(x_{t-1}) + \xi_t$, $y_t = g(x_t) + \zeta_t$
- 在**当前状态估计点**处计算 Jacobian: $F_t = \frac{\partial f}{\partial x}\big|_{\hat{x}_{t-1}}$, $G_t = \frac{\partial g}{\partial x}\big|_{\hat{x}_{t|t-1}}$
- 用线性化模型 $x_t \approx F_t x_{t-1} + c$, $y_t \approx G_t x_t + d$ 运行标准 KF
## 局限性
- Taylor 展开仅为一阶近似,强非线性下误差大
- 需要计算 Jacobian 矩阵(对高维系统代价高)
- [[nano-filter|NANO]] 证明 EKF 的线性化误差是 Gaussian 滤波次优性的根源,提出直接优化的替代方案
## 参考
- [[kalman-filter|Kalman Filter]]
- [[gaussian-filtering|Gaussian Filtering]]
- [[unscented-kalman-filter|UKF]]
- [[nano-filter|NANO Filter]]