theme: sudodoki/reveal-cleaver-theme title: PyData Berlin 2014 Talk - IPython and Sympy to Develop a Kalman Filter for Multisensor Data Fusion author: name: Thanks for attention! twitter: balzer82 url: http://balzer82.github.io/Kalman/ output: Presentation.html controls: true
--
~ Paul Balzer ~
--
> http://tinyurl.com/PyData-Kalman <
or
git clone https://github.com/balzer82/PyData-Berlin-2014-Kalman
~
Package Versions used:
Sympy 0.7.4.1
Numpy 1.8.1
Pandas 0.14.1
Matplotlib 1.3.1
IPython 2.1.0
Seaborn 0.3.1
~
Presentation made with cleaver
--
This talk will show you, how the IPython Notebook and Sympy helps to develop an optimal filter to fuse sensor information from different sources (e.g. IMU yawrate and GPS position) to get an optimal estimate.
--
A car is driving in the city. Shading > bad GPS Signal quality > Position signal is "jumping". But position estimate could be improved by speed and turn rate measurements.
--
- Sensor Noise
- Filter idea in 1D
- Multi-Dimensional Kalman Filter
- Extended Kalman Filter
--
Believe it or not: No sensor on the whole world is providing the 'real value'. And even if, you do not know, if it is the real value.
sensor value = real value + some error
Basic introduction to the theory behind the filter algorithm in just one dimension.
--
Now we are going multi-dimensional with matrices.
See Kalman-Filter-CA-Ball.ipynb
--
This is actually a real life example, which fuses GPS measurements and IMU measurements of a real vehicle.