ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | Q&A answers.ros.org

# EKF Slam problem with how to calculate Landmarks with Laser Data

Hi guys,

recently, I've been working on EKF SLAM problem on ROS using python, but I have some questions about the laser data usages.

Since my main matrix X is represented like X= [ x y theta x1 y1 x2 y2 .......... xN yN ] '

where x and y are robot's estimated pose, theta is orientation and x1, y1 represents the coordinates of first landmarks, while xN and yN represent coordinates of the N'th landmark.

But I'm not sure how to add these landmarks, because, with LaserData I have approx 200 + spots that Laser registers as an obstacle, so I'm not sure how to register it using only one x, y coordinate, but if I register every object with 200 or more landmarks, I will get insanely huge matrixes which will again, not be desirable or easy to calculate.

I've attached photo of my simulation so you can see what I mean. http://i66.tinypic.com/dlrvoo.png

Is there a way where I could for instance just define one wall with 1 meter as a landmark and somehow place it as a one x, y coordinate, but so that my robot recognizes entire line of wall as a landmark when it scans it.

TLDR; not sure how to represent landmarks in a simple way, when my laser data scans them in 200+ spots.

edit retag close merge delete

Sort by ยป oldest newest most voted

You have to extract features from LIDAR data to be able to run feature-based EKF SLAM. FLIRTlib and FalkoLib look like they could be useful for that.

Turns out there is even a catkinized version of flirtlib_ros available in Clearpath Robotics' flirtlib_ros fork.

more

I'm still not sure how, will this help me in initializing new landmarks in matrixes. Could you please try to explain it to me in more detail? Thank you in advance

( 2016-11-06 18:59:04 -0600 )edit

Sorry, I can't look into too much detail into this, but this might help; http://stackoverflow.com/questions/29... . Especially the this link in the comments: http://pastebin.com/m0SHQCqg

( 2016-11-07 06:04:01 -0600 )edit

Thank you for your help. I'll look into that :)

( 2016-11-07 06:29:31 -0600 )edit

## Stats

Seen: 966 times

Last updated: Nov 06 '16