ROS Answers: Open Source Q&A Forum - RSS feedhttps://answers.ros.org/questions/Open source question and answer forum written in Python and DjangoenROS Answers is licensed under Creative Commons Attribution 3.0Tue, 02 Dec 2014 02:49:08 -0600bfl - particle filterhttps://answers.ros.org/question/198639/bfl-particle-filter/Hello,
I'm trying to implement a particle filter using the BFL library from Orocos.
My problem is that my measurement system is multi-modal and in some cases non-analytical, with up to 30 measurements. For some measurements I have and expected value of 0, but for most of them I just know that they should be < 0, but I can't be sure in advance of the exact value. So, it s a bit different from the examples given in BFL website, or in the one from ROS, because they always know the value of the measurement they expect and then use a gaussian distribution to define to probability of this measurement.
I came up with the idea of applying every measurement value I get to a function that gives me an ouput between 0 and 1, and then multiply all these values into a final number (also between 0 and 1) that would be my probability for that measurement.
I'm not sure if this idea has any sense in for the measurement model. If not, can you suggest any way to handle this?
Thank very much!Mon, 01 Dec 2014 07:42:42 -0600https://answers.ros.org/question/198639/bfl-particle-filter/Answer by bvbdort for <p>Hello, </p>
<p>I'm trying to implement a particle filter using the BFL library from Orocos. </p>
<p>My problem is that my measurement system is multi-modal and in some cases non-analytical, with up to 30 measurements. For some measurements I have and expected value of 0, but for most of them I just know that they should be < 0, but I can't be sure in advance of the exact value. So, it s a bit different from the examples given in BFL website, or in the one from ROS, because they always know the value of the measurement they expect and then use a gaussian distribution to define to probability of this measurement. </p>
<p>I came up with the idea of applying every measurement value I get to a function that gives me an ouput between 0 and 1, and then multiply all these values into a final number (also between 0 and 1) that would be my probability for that measurement. </p>
<p>I'm not sure if this idea has any sense in for the measurement model. If not, can you suggest any way to handle this?</p>
<p>Thank very much!</p>
https://answers.ros.org/question/198639/bfl-particle-filter/?answer=198704#post-id-198704Your idea of function to convert measurement values to estimated values between 0 and 1 is correct.
But you should remember your estimates of 0 and 1 should represent least and higher likelihoods of particle.Tue, 02 Dec 2014 02:49:08 -0600https://answers.ros.org/question/198639/bfl-particle-filter/?answer=198704#post-id-198704