ModalAI Forum

1. Is there a log lin pipe field or debug flag that shows whether the throttle is currently engaged?
   Currently if you wish to have more robust logging you can run the commands via command line with the -d flag and check the output - you can also write some burner scripts capable of subscribing to the MPA itself and outputting what you need - lastly you can use voxl-logger which is capable or recording the data coming off the raw pipes.

2. Was the throttle behaviour changed after 0.6.1? If a newer voxl-open-vins-server fixes it, which SDK version should we move to?
   you should be pulling the most recent instance of it - you can either build it from source from gitlabl OR you can run apt-get upgrade and that will auto pull the most recent deb's from our server.

3. Given it is intermittent, is it safe to treat "dt == 33.3 ms" as a pre-flight check, or is there a more reliable condition we should test before an autonomous flight?
   I am a bit unsure what you mean treating this as a pre flight check - either openvins is feeding odometry and local setpoint positional data to px4 or it is not - if it IS even if its running slower that 33.3ms, then I would argue you are fine to fly.

Thanks!
Zach

https://www.loom.com/share/41699fad8b6842a6bf25c986c98c1c3b

Had it running on 8 minutes via voxl-uvc-server.

Can you confirm that you had good bandwidth the entire time? This is also over a microhard hat (although I have a wifi stick plugged into the usb port on here alongside the usb camera)

This sounds more like a networking issue than anything hardware or software related.

Ok lastly - can you please list out your OpenVINS package version? Previous versions of vins has a callback that alternates between processing and dropping frames when VIO is intiialized and its static jerk test permits throttling - so that first the 66.7 ms spacing you described. Restarting it would also bypass this throttle during init which is why it would go back to 30 hz

Can you put your voxl2 into performance mode and test the 15 vs 30hz please! voxl-configure-cpu perf

@jetson-nano I am testing this right now and am not seeing an issue - can you describe the following:

1. How long is the voxl on when you see this issue/how long has the stream been going?
2. Have you confirmed that voxl uvc camera library did not die and that is the cause?
3. Have you confirmed via the voxl streamer logs that the service did not die?

I am trying to understand which library is the source of the error in the mpa.

Thanks!

Hi! Planning on testing and getting into the weeds of this tomorrow to see if I can recreate and provide proper answers for this!

Zach

@arvind_g3aero - how many cameras are you running? Can you paste the logs exactly - it is hard to understand the issue without the logs.

As for snapshot - when you change the stream in qgc, the snapshot is probably not tied to the new rtsp steam, causing a fault on the backend - I can look into this at somepoint in the next fwe days.

As for 1.3.5 and 1.4.0 those SDK's do NOT support multi camera streaming into qgc.

Zach

@jon you can edit the dds via updating systemctl service file and point it towards the serial UART instead of a udp socket/port - the og looks like this:

[Unit]
Description=voxl-microdds-agent
SourcePath=/usr/bin/MicroXRCEAgent
After=network.target
After=multi-user.target

[Service]
User=root
Type=simple
ExecStart=/usr/bin/MicroXRCEAgent udp4 -p 8888

[Install]
WantedBy=multi-user.target

You would replace the -p with whatever flag is used for serial and then put /dev/ttyHSX (X number depends on which uart you are plugged into) - you will most likely need to put in a baudrate as well. This file is stored in /etc/systemd/systemd/voxl-microdds-agent.service.

^^^ This is the next best option outside of ADB IMO - nice suggestion @Alex-Kushleyev

@qt said in PX4 -> QGC connection through USB for VOXL2:

name

@bendraper I kind of redid exactly what you described above and ran into very similar issue in the sense that it basically rocked my usb-c port. All this to say, I think we should choose a different route - is it possible instead to route the packets via adb over the USB-C port instead to basically get the same result? Point the software to the serial connection and leverage ADB to pull the packets that are forwarded? Basically creating our own sudo rndis connection almost.

Zach

Hi All - gonna spend a bit of time looking into this today so I can get some context on the problem.

Zach

@voxltester said in Microdds not working with Microhard modem:

Is there any one who is an expert on the microhard and network setting?

You will need to reach out to microhard for config setup on that front.

@Nitin-Varma-Vegesna said in Apriltag relocalization not relocalizing?:

@Nitin-Varma-Vegesna Is there any release that is known to work for April Tag Detection?

https://gitlab.com/voxl-public/voxl-sdk/services/voxl-tag-detector

The dev branch is the most recent - I have to check with the engineering team on this as I am unsure the priority of april tag detection and positioning based off it. Will relay back

https://www.loom.com/share/bf52e252ab09444bb366f265a3f36dc5

Alright take a look at this loom please - it might help point you in the right direction in terms of training your model.

Zach

Let me try and train a custom model and run a loom on it in the next few days and get that over to you showing how I do it!

@voxltester said in Microdds not working with Microhard modem:

Do you have any ideas on how to resolve this?

There might be a dependency in the service file - I would recommend taking a look at the file in /etc/systemd/system/voxl-microdds-agent.service and see if there are any dependencies on having WLAN devices plugged in - I am under the impression there is not that dependency, but there is a conflict on which interface you want the dds to start on - so ideally you can add in which interface you wish to leverage into the service file.

Zach

@svempati I was able to use an open source training set and then leverage the docs to make my own custom yolov8 model capable of running on the voxl2 - do you want to provide to me the actual dataset and I can try creating the model and training it on a TPU?

@svempati said in No detections when running custom YOLOv8 model on voxl-tflite-server:

WARNING: Unknown model type provided! Defaulting post-process to object detection.

@svempati I will try recreating this issue today and get back to you!

@Peter-Lingås-0 I do not have an answer for any global shutter cameras but what I can tell you is that the AR0144 camera can be used for anything as well not just VIO - the frames being piped out go into the software dedicated for OVINS but it can also be piped anywhere else due to the pub/sub architecutre, so you should have no problem piping that into your computer vision software either.

I would recommend doing a barometer calibration as well via voxl-calibrate-barometer that tends to solve a good chunk of the altitude issues.