You will have to study PX4 architecture and work through the code to create a new type of GPS input Did you go through the VIO troubleshooting steps? https://docs.modalai.com/flying-with-vio/#troubleshooting-vio

nvm, solved it. Had to enable and run the service voxl-mavlink-server first, then do the same with the voxl-vision-px4 service. I'll leave this here for posterity

Hi, While we so technically support mapping using stereo we almost always use the tof sensor ourselves since it's much more accurate at higher resolutions. The default for our voxl-mapper project uses the tof sensor for mapping with option to enable stereo if that's what you want to use. We generally just use stereo for collision prevention since it's pretty reliable at detecting if something is in front of it or not but can struggle to populate a detailed full-fov image in realtime.

Thanks Eric, I'll look into the legacy mode

Sorry it's mikrobus.

@Steve-Turner said in Seeker drone "backtracking" when it stops to hover: I don't work for Modal, so take this with a grain... MPC_POS_MODE controls some of this behavior. https://docs.px4.io/v1.12/en/advanced_config/parameter_reference.html has a description and you can try the other options to see how that influences the behavior. https://docs.px4.io/v1.9.0/en/config_mc/advanced_mc_position_tuning.html Should help with some of the tuning parameters that control position behavior. Additionally, the PID loop for position might need some tuning if you want it to stop more aggressively. Thanks for the suggestion but I'm quite certain it has nothing to do with MPC_POS_MODE nor the tuning of the velocity/position controller. The phenomenon I described only happens sometimes - especially when VIO quality is low for that stretch of journey just before the drone comes to a stop

@yitakabe Can you post either a screenshot or copy/paste the error you're seeing?

OK. We don't have a documented way, but VOXL 2 is a Linux computer. So you could use some form of networking: ROS or other. There are also UARTs, etc

QVIO does not perform SLAM, only visual tracking to aid in stability by tracking the velocity vectors of key points through a camera. https://gitlab.com/voxl-public/voxl-sdk/services/voxl-mapper performs mapping and planning. Data can be imported and exported through VOXL Portal with "Load Map" / "Save Map"

Interesting! I'll keep watching this thread for when you commit a proper fix in voxl-modem. For the meantime, my hack is working for me. Thanks!

@ahmed-mustahid Unfortunately the "QUSB_BULK" mode that your VOXL is enumerating in cannot be fixed by our unbricking process and will have to be sent in for an RMA. See here for info on the RMA process: https://www.modalai.com/pages/rma

@Maxim-Temp That image is also still in test.

@Eric-Katzfey Hi Eric. The jpeg compression imposes a lot of load on the CPU, especially we're using the hires camera. Is there any way you can make the HW jpeg encoding available for us to test it? Thanks!

@tom Ok, thanks!

@Adrian-Hidalgo I get it. But from the moment I am giving throttle the vehile starts wobbling and it increase. I am unable to fly it for a second.

@Adrian-Hidalgo @Dobry-Kolacz Thanks for your reply, so the drone uses QVIO to control the altitude (I think it uses barometer or something else) which perfectly explains why the altitude sometimes drifts as our floors are a bit lacking in texture. I'll try adding some texture to the floor and give it a try. Thanks.

Yes, the service doesn't need to be enabled on voxl/voxl2 (the configure-mpa script will enable the correct imu server for whichever platform you're on)