ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
1

Is there any black box testing tool in ROS?

asked 2018-06-08 23:37:11 -0500

Changkun gravatar image

updated 2018-06-09 00:12:01 -0500

Hi, I want to test my SLAM system whose input is sensor msgs(camera, imu ...) and the output is the map and pose.

I'd like to do black box testing on the system(several nodes) , by giving the input and checking if the output right. I think I can record a rosbag with the inputs and then insert groundtruth at specific timestamp into it. By checking the output and the groundtruth of the same timestamp, I can see if the system run well.

Is there any tool or framework to do this kind of integrating testing on system?

edit retag flag offensive close merge delete

Comments

1

I think the title of your post is too generic: you're specifically asking for a testing tool for SLAM algorithms. As there could be some stochastic elements in there (or even just plain optimisations), output may not always be 1-on-1 identical between runs. Simple == output with groundtruth ..

gvdhoorn gravatar image gvdhoorn  ( 2018-06-09 05:52:36 -0500 )edit

.. would not be sufficient. A more qualitative approach such as the one suggested by @NEngelhard would make more sense.

May I suggest to, at the very least, mention the SLAM aspect / context in the topic title?

gvdhoorn gravatar image gvdhoorn  ( 2018-06-09 05:53:29 -0500 )edit

Ok. I just want to test SLAM. There seems no general system to do black box testing on nodes.

Changkun gravatar image Changkun  ( 2018-06-10 21:53:47 -0500 )edit

2 Answers

Sort by ยป oldest newest most voted
0

answered 2018-06-09 11:00:10 -0500

ahendrix gravatar image

The gmapping test suite does something similar. There isn't any documentation about it; you'll just have to read the code to understand how it works: https://github.com/ros-perception/sla...

edit flag offensive delete link more

Comments

It compares the map with groundtruth pixel by pixel, that makes sense.

Changkun gravatar image Changkun  ( 2018-06-10 22:10:54 -0500 )edit
0

answered 2018-06-09 02:53:18 -0500

NEngelhard gravatar image

I don't know any general test system, but you could have a look at the RGBD-SLAM Dataset that was developed in Freiburg and Munich: rgbd-dataset The Evaluation Tools take an estimated trajectory and the groundtruth as input an generate several error metrics and plots.

edit flag offensive delete link more

Comments

I see. Thank u.

Changkun gravatar image Changkun  ( 2018-06-10 22:08:32 -0500 )edit

Question Tools

3 followers

Stats

Asked: 2018-06-08 23:37:11 -0500

Seen: 207 times

Last updated: Jun 09 '18