Logs not started on all CFs in swarm

[krichardsson]

Hi!

I'm trying to write a simple test for our test lab to verify the broadcast functionality of the radio. I have added a simple function to the firmware where I have connected a parameter to a log, that is the value written to the parameter (system.testLogParam) will be logged by the log (sys.testLogParam). The idea is to set the param with a broadcast and then verify the result with the log.

It sort of works, but I see a weird behavior. We have 10 Roadrunners in the swarm and the log seems to be set up only on the 5 last members in the swarm. This seems to be true also if I remove a couple of members, and with 5 members all of them get the log. Is this an expected behavior or am I doing something wrong?
I can see on the LEDs that only the 5 last RR are transmitting.

The relevant code looks something like this:
In the .launch file

      genericLogTopics: ["log1"]
      genericLogTopicFrequencies: [1000]
      genericLogTopic_log1_Variables: ["sys.testLogParam"]
      # ...
      enable_logging: True

In the python test

        def log_callback(value):
            rospy.loginfo(value)

        # Works for nr 7 but I get no logging for nr 0
        cf_id_to_get_logs_from = 7
        rospy.Subscriber('/cf' + str(swarm.allcfs.crazyflies[cf_id_to_get_logs_from].id) + '/log1', GenericLogData, log_callback, queue_size=1)

        swarm.allcfs.setParam('system/testLogParam', value)
        rospy.sleep(10)

rostopic list reports the log1 topic on all members.

Any help is appreciated!

[whoenig]

Crazyswarm1 is not very well behaved when using logging for more than 1 CF (it was essentially only tested for debugging issues with a single CF). One thing to be careful about is the "genericLogTopicFrequencies" - from the name one would guess that this is Hz, but in reality it's the period in 10ms increments, so in this case you would get new data every 10s, which is perhaps not what you expect. There are also issues with fairness - it's entirely possible that some CFs are "starved" if you use a single radio.

In theory, Crazyswarm2 is better in all these regards, but I think it's not set up in your test lab yet.

[krichardsson]

That's the feeling I got :-)

The main purpose was to verify that broadcast works on the Crazyradio2. It has been tested manually and I think I'll wait with the automated test until we switch to CS2.

Thanks!

[devbf]

Hello @krichardsson,

I have a similiar problem at the moment, I want to get the log1 topic via ROS.
The problem is that I can´t find the import which is needed for the GenericLogData datatype. I´ve searched through the entire crazyswarm repository and did some general searches, but the few code snippets I´ve found all show no imports.

I´d be glad if you could tell me the import directive you´ve used to get the GenericLogData datatype.

Thanks in advance

Flo

[devbf]

After endless tries and a long conversation with ChatGPT I finally found the solution:

1. I had to do a "catkin_make" in the crazyswarm/ros_ws directory
2. Then I could use the following import command: from crazyswarm.msg import GenericLogData
3. Even though the import statement is underlined in yellow color by Python, it works

I hope if someone in the future stumbles upon the same problem, this will help.