I heard that time synchronization is extremely critical for VIO systems.
Yes, but it is not really easy to archive it.
In my experience. you at least need to
do you think switching to a global shutter camera like the OV9281 would significantly improve VIO performance
Yes, I would strongly recommand use a global shutter camera
thank you,
I’m using ros2, and there has been issues with ros2_version of ego_planner. GitHub - ZJU-FAST-Lab/ego-planner-swarm at ros2_version
I have created an issue Issue with using planner on real drone · Issue #82 · ZJU-FAST-Lab/ego-planner-swarm · GitHub , but no one seems to be active.
Could you please review and let me know where I’m going wrong?
The launch file is setup exactly like yours ros1 version ego-planner/src/planner/plan_manage/launch/vins.launch at pi4 · chobitsfan/ego-planner · GitHub.
Is the planner an open-loop system, expecting that the drone will follow the velocity command?
Hello, i have the same issue. Have you figured it out?
did you use hardware timestamp to match camera - imu time sync?
In recently projects, I use oak-d lite because its camera & imu timestamp came from the same counter
Hi, @chobitsfan. First off very cool stuff. Now to my question: have you tried to fly your system outdoors? I tried it off for a bit indoors, and it worked fine (although I didn’t fly too much indoors, as it’s a bit scary to fly 13 inch drone indoors), but when i took it out on street, it appears tracking starts to fail when it’s encountering a scene with variable lighting conditions (an empty street with trees from each side, the street is well lit, while the trees part is much darker), it worked fine when i was simply flying along the street, but failed miserably when i yaw towards trees, which is why i though it could be the tracking issue. I tried it with both automatic and manual exposure controls, and the result is more or less the same. do you know what could be done to fix this behavior? Or could it perhaps tie to my custom-built ardupilot (i built an ardupilot 4.6.3, but configured it so it would include visual odometry and highres_imu messages, as it doesn’t support it if built normally)?
Hi @IssatayS
It seems monocular VIO sometimes seems to drift away while rotating. I found stereo VIO is more stable while rotating. Would you mind trying Luxonis OAK-D — Copter documentation ?
May I know where the readme.md for mavlink-udp-proxy is?
use this branch GitHub - chobitsfan/mavlink-udp-proxy at apm_wiki