@Ed-Sutter Thanks! Always better to start a new thread for a different question.

Here is the log: -- Logs begin at Thu 1970-01-01 00:01:35 UTC, end at Thu 1970-01-01 00:03:25 UT C. -- Jan 01 00:01:37 apq8096 systemd[1]: Started voxl-vision-px4. Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: ERROR validating VIO data receiv ed through pipe: read partial packet Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: read 156 bytes, but it should be a multiple of 324 Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: ERROR validating VIO data receiv ed through pipe: read partial packet Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: read 156 bytes, but it should be a multiple of 324 Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: ERROR validating VIO data receiv ed through pipe: read partial packet Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: read 156 bytes, but it should be a multiple of 324 Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: ERROR validating VIO data receiv ed through pipe: read partial packet Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: read 156 bytes, but it should be a multiple of 324 Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: ERROR validating VIO data receiv ed through pipe: read partial packet Jan 01 00:01:38 apq8096 voxl-vision-px4[2018]: read 156 bytes, but it should be a multiple of 324

Thank you!

Yes it should work -- I haven't explicitly tried pointing directly downwards myself. https://docs.modalai.com/flying-without-gps/ has some insight and a nice video about the subject of using VIO outdoors. It does work at 45* reasonably well especially with enough altitude to pick out features.

Hi, The default framerate of 15 hz will limit the sensor to around 2-3 meters, lowering this to 5hz will allow you to see the full range of the sensor. As for the IR image, it's meant to help debug what the sensor sees, this isn't a normal camera with exposure/gain so the IR image will represent what the sensor actually sees, not necessarily in the most human-viewable format (Note that we often find the confidence image looks better to our eyes for demos).

@darwin-keem, It looks like voxl-mapper failed to reach your goal point, resulting in it sending over the partial trajectory it solved for viewing but not giving you the option to follow the path. You can try giving the RRT solver another second or so by increasing the rrt_max_runtime_nanoseconds param, and it should be able to fully plan to your endpoint. The path generation code has been rewritten and thoroughly optimized in the latest dev package, so if you are still having issues with this planner it may be worth checking out some of the updates on our dev repo.

https://forum.modalai.com/topic/1328/camera-servers-docker-build-royale-331-spectre-4-7 may have some answers for you on voxl-camera-server which is something ModalAI is working on WRT the build.

The camera server is working with another set of cameras after I reflashed the voxl 2. This might be a hardware issue. Thank you.

Hi, We've published an RB5 platform release that uses the entire VOXL SDK stack instead of the rb5 software suite. This supports using voxl-camera-server and voxl-streamer as you would on a voxl or voxl2. This release can be found here

Hi @Anton-Gereles , Can you please elaborate on your platform and specific problems you are having? I'm sure our Software team will want to know that to respond appropriately. That being said, our components and hardware elements are all at least rated for 0-70C, with many of them exceeding that. However, be mindful that is a non-condensing requirement. If you are operating in extremes (we have many customers that do), you need to be mindful of condensation risks. Rapid change in temp from a "room temp" building into an outdoor setting with high humidity and much colder temps (and especially vice versa) can create condensation. In these types of environments, you may need to provide some type of enclosure to protect the electronics against this build-up. Also, our products are not dust and particle ingress resistant as a default, again this would be something we leave our customers to define for their industrial/mechanical design requirements for an enclosure. Hope that helps and if you do provide more failure info, our SW team could respond. Thanks!

Hi, VOXL comes with FastCV installed, the 2 headers can be found in /usr/include/fastcv/ and the libs are in /usr/lib/ : voxl:/$ find /usr | grep fastcv /usr/include/fastcv /usr/include/fastcv/fastcv.h /usr/include/fastcv/fastcvExt.h /usr/lib/libfastcvadsp_stub.so /usr/lib/libfastcvopt.so /usr/lib/rfsa/adsp/libfastcvadsp.so /usr/lib/rfsa/adsp/libfastcvadsp_skel.so

Hi, You can definitely try it but the QVIO algorithm benefits A LOT from having a wide field of view on the camera and the hires cam has a fairly narrow field of view, especially when compared to the fisheye tracking camera. There are no hard restrictions to trying this but know that performance will almost certainly be worse.

@vivek-rk For anyone looking for a quick fix, you can use the voxl rtsp server for the hires only, for MPA support use the add-mpa-support branch

@Eric-Katzfey Yea... good thought... I'll look into that... Thx