Ask Your Question
4

Rostopic Hz inaccurate?

asked 2015-04-02 04:35:04 -0600

Sannoso gravatar image

Hello everyone.

Sometimes when I measure the frequency of messages in a topic, the numbers printed don't make sense.

For example:

rostopic hz /robot/head/head_state

gives:

subscribed to [/robot/head/head_state]     
average rate: 100.000
        min: 0.007s max: 0.015s std dev: 0.00097s window: 96
    average rate: 100.000
        min: 0.004s max: 0.016s std dev: 0.00111s window: 192
    average rate: 100.000
        min: 0.004s max: 0.016s std dev: 0.00100s window: 289
    average rate: 100.000
        min: 0.004s max: 0.016s std dev: 0.00112s window: 386
    average rate: 100.000
        min: 0.004s max: 0.016s std dev: 0.00105s window: 483
    average rate: 100.000
        min: 0.004s max: 0.017s std dev: 0.00109s window: 581
    average rate: 100.000
        min: 0.004s max: 0.017s std dev: 0.00116s window: 678
    average rate: 100.000
        min: 0.004s max: 0.017s std dev: 0.00115s window: 776

So it measures 776 messages in 8 seconds and still tells me that the average rate is 100.000? My calculater tells me that there are 776 messages over 8 seconds -> = 97Hz. Why aren't the numbers in my terminal not corresponding with this?

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted
3

answered 2015-04-07 07:08:48 -0600

Sannoso gravatar image

After a suggestion of tfoote, I found out that the bias was caused by simulation time. (I did not know there was such a thing). The added picture is a part of the gazebo simulator that I ran today. For a reason that I don't know, there is a real time factor of 0.97 and not 1.00. This explains the 97Hz.

image description

My conclusion is that the Rostopic Hz function prints every REAL second the amount of messages observed, but when calculating the average rate it compensates for the Real Time FActor, so the outcome is 100 Hz.

Thank you for your help tfoote.

edit flag offensive delete link more
3

answered 2015-04-03 12:16:29 -0600

tfoote gravatar image

I would guess that your assumption of sampling being conducted for exactly 8 seconds might be off by 0.24 seconds due to startup connection times etc.

Feel free to dig into the source at: https://github.com/ros/ros_comm/blob/...

edit flag offensive delete link more

Comments

Thank you for responding. But if the bias would be caused by startup connection times, then why does it occur not only in the first second, but in all seconds after that aswel? I mean, for example 386 - 289 = 97. 97 messages detected in that second, but still it prints an average rate 100 Hz?

Sannoso gravatar imageSannoso ( 2015-04-05 10:44:55 -0600 )edit

Can you provide instructions for how to reproduce your output?

tfoote gravatar imagetfoote ( 2015-04-05 13:06:58 -0600 )edit

Unfortunatly I can't because a piece of software that is being used is from a private repository which I am not the owner of and I haven't been able to reproduce the output with something else. If I find out what the answer is I will post. Thank you for your effort.

Sannoso gravatar imageSannoso ( 2015-04-07 01:59:34 -0600 )edit

One other question, are you running with simulated time?

tfoote gravatar imagetfoote ( 2015-04-07 02:37:47 -0600 )edit

Ah. I did not know of its existence. This solves my problem. Thank you. I wrote an answer for it.

Sannoso gravatar imageSannoso ( 2015-04-07 07:10:33 -0600 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower

Stats

Asked: 2015-04-02 04:34:22 -0600

Seen: 3,032 times

Last updated: Apr 07 '15