Hello,
I would like to apply BFL library for my current project.
Short description:
In our laboratory we have MEMS gyro, which is attached to industry robot with 6 independent axis for calibration purposes. For communication with robot and gyro using REAL-time support Xenomai. I have got very litle sample rate about 2ms. Because the output of my gyro is loaded with white noise I would like to apply
Kalman filter, but I dont know is it possible to do some change in the BFL project for our purposes.
I studied example in BFL project and seems to be very suitable for our project. But I dont know how can I set the sample rate for our measurement. Did you solve similar problem with filtering the output of gyro?
Pavel.
Using Extended Kalman Filter for output of gyroscope and acelera
On Thu, 28 Jun 2007, pavel.pirkl wrote:
> Hello, I would like to apply BFL library for my current project.
>
> Short description: In our laboratory we have MEMS gyro, which is attached to
> industry robot with 6 independent axis for calibration purposes. For
> communication with robot and gyro using REAL-time support Xenomai. I have got
> very litle sample rate about 2ms. Because the output of my gyro is loaded
> with white noise I would like to apply Kalman filter, but I dont know is it
> possible to do some change in the BFL project for our purposes.
>
> I studied example in BFL project and seems to be very suitable for our
> project. But I dont know how can I set the sample rate for our measurement.
> Did you solve similar problem with filtering the output of gyro?
>
Of course, we continuously do things like that: using BFL in realtime
estimation loops :-)
BFL is "just" a library, so the _timing_ is not being done in BFL, but that
is the responsibility of your realtime process. So, just _call_ the BFL
functions in the loop where you need them; if the hardware is performant
enough, hard realtime will be possible.
Herman
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
Using Extended Kalman Filter for output of gyroscope and acelera
Thanks Herman
The Xenomai offer hard real-time.
So I can do loop for example following way:
number_sample=1000;
sample_time=0.100;
//set rt_task with periodick behaviour to given sample time
rt_task_set_periodick(&rt_task, TM_NOW, sample_time);
for (i=0; i
//wait rest of time of sample_rate
rt_task_wait_periodick();
//get actual measurement of the output of gyroscope
measurement=gyro.Measure();
//the compute of this procedure must not overlap sample_time
filter.Update(&sys_model,input,&meas_model,measurement);
//get estimate value of sytem model: angle_rate, angle, ..
posterior=filter.PostGet();
estimate_value=posterior->ExpectedValueGet();
//save result of estimation to file for example
save_result(estimate_value);
}
Do you have some experience with setting of parameters of system model, measurement model for gyroscope or accelerometer?
I am solving inertial system for mobile robot which is based on the gyroscope and accelerometer.
In future I would like to use this system with data fusion with odometry or GPS.
Is it possible to use the BFL for data fusion?
Pavel.
Using Extended Kalman Filter for output of gyroscope and acelera
On Thu, 28 Jun 2007, pavel.pirkl wrote:
> The Xenomai offer hard real-time.
> So I can do loop for example following way:
>
> number_sample=1000;
> sample_time=0.100;
> //set rt_task with periodick behaviour to given sample time
> rt_task_set_periodick(&rt_task, TM_NOW, sample_time);
> for (i=0; iExpectedValueGet();
> //save result of estimation to file for example
> save_result(estimate_value);
> }
It would not be wise to store results in a file while you are in realtime
mode! (And the for "loop" in your example would not even compile.)
> Do you have some experience with setting of parameters of system model,
> measurement model for gyroscope or accelerometer?
No.
> I am solving inertial
> system for mobile robot which is based on the gyroscope and accelerometer.
> In future I would like to use this system with data fusion with odometry or
> GPS. Is it possible to use the BFL for data fusion?
Of course... You don't seem to have awful lot of background about the
things you are doing :-) If I were you, I would rather study some Bayesian
probability before trying to use code that implements it...
Herman
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm