Skip to content

Latest commit

 

History

History
106 lines (63 loc) · 4.38 KB

README.md

File metadata and controls

106 lines (63 loc) · 4.38 KB

Can you outrun the Big Bad Kalman filter ?

Some linear, extended and unscented movement tracking Kalman filters, with a fun twist

View Object Tracking via Sensor Fusion on File Exchange

image

Run ObjectTracker.m and make sure all files are in the same directory. Set your scenarios using the dropdowns.

Press Play and enjoy :-)

Go for Developer Mode if you want to generate your own custom data and play around with the trackers:

Model Parameters Filter Tuning Extra Sensor
1 2 3

Note You can control the Seal if you own an Arduino + MPU IMU sensor suite, this is how it works.

To achieve this, you may choose Command Driven instead of Simulation for the Running Mode.

Demos

Noob level: defeat the linear Kalman filter

The Shark can only chase you in a linear fashion

Test each of your runs:

image

image

Experienced: defeat the extended Kalman filter

The Shark is getting help from a Seagull, who acts like a sensor for detecting your non-linear movements

Measure your performances:

image

image

Note You can trick the shark by moving fast in a non-linear manner

This way you can make the filter diverge due to wrong partial derivative computation

Legendary: defeat the unscented Kalman filter

No more linear covariance transforms, the Shark has unlocked the Unscented Transform ability

And see how far your can get:

image

image

How this madness was designed:

engineered:

and programmed:

with the following workflow:

and if you made it this far...

here is the whole thing explained in detail (Vampire language):

OneFilterToRuleThemAll.pdf