0
$\begingroup$

Rosanswers logo

Still setting up ekf_localization_node here -- I'm playing back and processing my bag, but getting this error, whether I run from a lauch file with params or with no params from rosrun,

$ rosrun robot_localization ekf_localization_node

ekf_localization_node: ../nptl/pthread_mutex_lock.c:350: __pthread_mutex_lock_full: Assertion `(-(e)) != 3 || !robust' failed.
Aborted (core dumped)

The error only occurs when by bag is playing, otherwise the node sits silently. There's no log file or rosout messages so I'm not sure how to locate the problem, any suggestions ?

FURTHER DETAILS:

System info: PARAMETERS

  • /rosdistro: indigo * /rosversion: 1.11.10 on Xubuntu 14.04 LTS.

Sample geometry/gps message,

header: 
  seq: 200
  stamp: 
    secs: 1431535613
    nsecs: 57244062
  frame_id: ''
child_frame_id: ''
pose: 
  pose: 
    position: 
      x: -2.90984643623
      y: -3.69588956097
      z: -3.51702306711
    orientation: 
      x: 0.0
      y: 0.0
      z: 0.0
      w: 1.0
  covariance: [1.218219225942454, 0.006998948102184068, 0.17239271800594577, 0.0, 0.0, 0.0, 0.006998948102183846, 1.2159598403645349, 0.14679821372057594, 0.0, 0.0, 0.0, 0.17239271800594574, 0.14679821372057594, 4.825820933693012, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
twist: 
  twist: 
    linear: 
      x: 0.0
      y: 0.0
      z: 0.0
    angular: 
      x: 0.0
      y: 0.0
      z: 0.0
  covariance: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]

Sample /imu message:

header: 
  seq: 105
  stamp: 
    secs: 1431535415
    nsecs: 50663948
  frame_id: base_imu_link
orientation: 
  x: 0.026096916629
  y: -0.0692553241391
  z: -0.248199192676
  w: 0.96587774163
orientation_covariance: [0.0025, 0.0, 0.0, 0.0, 0.0025, 0.0, 0.0, 0.0, 0.0025]
angular_velocity: 
  x: -0.09
  y: -0.04
  z: 0.02
angular_velocity_covariance: [0.02, 0.0, 0.0, 0.0, 0.02, 0.0, 0.0, 0.0, 0.02]
linear_acceleration: 
  x: 0.666884609375
  y: 0.117595390625
  z: 9.37468921875
linear_acceleration_covariance: [0.04, 0.0, 0.0, 0.0, 0.04, 0.0, 0.0, 0.0, 0.04]

Launch file,

<launch>

  <node pkg="robot_localization" type="ekf_localization_node" output="screen"  name="ekf_localization_node">

   <param name="odom0" value="odometry/gps"/> 

   <param name="odom0_differential" value="false"/>

   <rosparam param="odom0_config">[true,  true,  false, 
                                    false,  false,  false,
                                    false,  false, false, 
                                    false, false, false,
                                    false, false, false]</rosparam>
  
   <param name="imu0" value="imu" />

   <rosparam param="imu0_config">[false,  false,  false,
                        false,  false,  true,
                        false,  false, false,
                        false, false, true,
                        true, false, false]</rosparam>

   <param name="imu0_differential" value="true"/>

   <param name="imu0_remove_gravitational_acceleration" value="true"/>
    
  </node>
</launch>

Update:

To try to isolate the cause, I tried recording my data into smaller bags, and found that it's the presence of tf data in the bag that causes it. ie, this leads to the error,

rosbag record assemble clock imu /imu_node/parameter_descriptions /imu_node/parameter_updates /odometry/gps tf

but this does not:

rosbag record assemble clock imu /imu_node/parameter_descriptions /imu_node/parameter_updates /odometry/gps

so what could be going on? AFAIK the ekf node isn't using tf info, so why does the error depend on it?

I also get the same error swiching efk to ukf.

Originally posted by charles.fox on ROS Answers with karma: 120 on 2015-05-22

Post score: 0

Original comments

Comment by Tom Moore on 2015-05-22:
Can you share one of the bag files? Have you tried just using the IMU or just the GPS data? I'm interested in your frame-less GPS data, though that shouldn't cause it to crash. Also, did you compile from source? If so. which branch?

Comment by charles.fox on 2015-05-22:
Hi Tom, great to see you here! I've put a (tiny) bag and my launch file here for you, https://dl.dropboxusercontent.com/u/1481226/a.bag https://dl.dropboxusercontent.com/u/1481226/t.launch The crash happens even if I have IMU only, or GPS only, or comment out both of them. I installed from apt-get

Comment by charles.fox on 2015-05-22:
still have same error after looking at the empty gps frames. My navsat_transform_node is taking its frame from my own command-based odometry node . This was sending empty frames so I filled them in, parent=odom_motor, child=base_link. odometry/gps now has the same parent but still empty child?

Comment by Tom Moore on 2015-05-22:
Yeah, see my answer below. Let's first move away from the installed version and go to source for now. Then we can try to work through these issues one-at-a-time.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Rosanswers logo

Ok, so a few things:

  1. I am unable to reproduce the behavior you describe with the latest source (which will hopefully be getting released imminently). The node starts just fine either way and doesn't crash. There could be an incompatibility between my node and an updated dependency in another package. Regardless, the error you're seeing involves threading, and I don't explicitly use threading anywhere, which means it likely is an interaction with a dependency.
  2. Your IMU is given in the base_imu_link, but you have not specified a transform from base_imu_link->base_link. See this section of the wiki.
  3. In your launch file, you are not measuring nearly enough variables. If you are operating in 2D, then set two_d_mode to true. Otherwise, you need to measure enough variables to keep your covariances from exploding. See this page. You might also try fusing your commanded velocities.
  4. You appear to have a transform from odom_motor_cmds to base_link, but it appears to be changing. If you have some node that is outputting motor commands to the robot, I would suggest issuing them as geometry_msgs/TwistWithCovarianceStamped messages. That way, you can fuse them with the state estimate.

Originally posted by Tom Moore with karma: 13689 on 2015-05-22

This answer was ACCEPTED on the original site

Post score: 0

Original comments

Comment by charles.fox on 2015-05-22:
Thanks Tom. I fixed base_imu_link and the error remains; and I'm already in two_d_mode. odom_motor_cmds computed by my own controller, I'll try switching that to TwistWithCovarianceStamped as the problem is something to do with tf. And try latest source.

Comment by charles.fox on 2015-05-24:
Seems to be working now I upgraded to latest source (2.1.7), thanks very much for the help Tom!

Comment by Tom Moore on 2015-05-24:
Odd. 2.1.7 was the release you had already. You just compiled from source instead. Perhaps there was some strange build flag or incompatible dependency. Oh well. In any case, you should update to 2.2.0 (literally just released).

Comment by charles.fox on 2015-05-24:
I'm cloning from git pasting the URL from github, this gives 2.1.7: https://github.com/cra-ros-pkg/robot_localization.git but git checkout -b remotes/oriin/indigo-devel; git pull origin indigo-devel fxes package.xml (only) to 2.2.0. So maybe this /is/ 2.2.0 but package.xml isn't updated in master?

Comment by charles.fox on 2015-05-24:
(done a diff on the that clone and a clean clone with git clone -b indigo-devel https://github.com/cra-ros-pkg/robot_localization.git
, the only difference is the version number string in package.xml, so looks like the master has 2.2.0 but with the old package.xml, saying its 2.1.7?)

Comment by Tom Moore on 2015-05-24:
Yeah, I don't update the package.xml version until I do a release, so even though your package.xml version said 1.1.7, it was the most recent code.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.