Kalman Filter

http://en.wikipedia.org/wiki/Kalman_filter

我的NXT Sensor終於都可以把資料傳到PC上了
因此有空就會去找一些有關Auto Navigation的Web
好像大家都會提到這個 Kalman Filter.
基本概念應該都了解了
我覺得GPS Tracker應該都要Implement.
而且大部分的Sensor Input之後也應該要
先用Kalman Filter處理過
不只是Signal品質會比較好 而且我對
是不是能從Prediction中找出一點東西來利用很感興趣

Comments

[Fred]

Kalman filter is huge in this town (Hutsville) as scentists here use it from controling self-navigating robots/vehicles, predicting the trajectory of a storm or flow of rivers during flood and even the prices of traded commodities and stocks. I actually met Rudy Kalman once when he came to Huntsville gave us a talk a few years ago. He is a hero figure here perhaps second only to Werner Von Bruan (a true German-American Rocket scientist).

Kalman is a mathematical tool which can only be as good as the underline model (state machine) of the system. If your model is incorrect, you may get completely erroneous prediction even when you calculated the "optimal" solution based on Kalman filtering.

[Tze-Chien Chu]

1.Can you recommand some book for me to study ?
2.The model is mean the Signal I received or the Noise Model ? Or the How to Predict ?
3.Actually I got a more stupid but simple way. It is not for signal filter but for decision making
a.I use Moving Average as Prediction.(MA)
b.The value I care about is not the signal but the DeltaSignal/MA.
DeltaSigal is S(n)-S(n-1)
Is that make sense ?