hey there! The position and velocity are states we are interested in and are not input to the system strictly; they are included at the priors, but the the acceleration is the input to the system, it's added to the prior in a sense. you absolutely can add it to the state vector, but it's all about what you're trying to model. here, i was just trying to follow the state position in a simple way. but yes you can add it! have a good new year.
i really like the way you elaborate, thanks for sharing your knowledege
yup, but in just this example. again, you can definitely add it!
Hi Dave,
u are using linear equations for the physical model to predict the next step. But isn´t the hexbug a non linear system? shouldn´t u use the extended Kalman-Filter?
excellent, especially setting up the matrices
Great Tutorial.
I'll try it on matlab.
Can you post a link to the video u're using?
thank! more soon! hmm..not sure exactly what you mean but as long as the prediction and measure are different the actual gain will be non-zero and likely change. shoot me an email on my website and we can talk more. can't wait to see your work!
Hey, what i don't understand is what value to use for acceleration. Since you treated acceleration as an input/control to the system when in fact we have no control over the system. Should we assign acceleration to be a constant? and if so how will this constant affect the behaviour of the tracking?
thx!
This is so cool !!
great job!
hi do you have code for input data from a sensor e.g. a kinect sensor, instead of the input data being from picture file?? for instance inputting the x,y,z co-ordinates into the code instead
Hi ,I found a -74% coupon to Learn Matlab Udemy course
Couponcode: 2016ML25
www.udemy.com/learn-matlab/?couponCode=2016ML25
check out the link to my website! it's there with a whole lot more to play around with! :)
Singular is "matrix", not "matrice". Really nice explanation, but I found myself involuntarily wincing each time you say "matrice" :-)
the measuring error should be "x & y", but the "y" looks like "z"
where I can found the video
Cool!
Ive been trying to implement a kalman filter so i can track an air hockey puck nicely ! Been banging my head against it for a week only to realise i didnt zero out the xy relationships for my process noise ! Your video saved me