-
Notifications
You must be signed in to change notification settings - Fork 4.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
T265 is loosing position under vibrations #4176
Comments
Vibration may be more of a problem if you are using SLAM than it is with depth sensing on the 400 Series cameras (as the 400 Series has very robust stereo vision algorithms). On the T265, 'wheel odometry' is used as a stabilizing mechanism. Wheel odometry is not likely to be useful on a flying drone though if there is no turning of a robot's wheels involved. There is an example on YouTube of T265 being used with a flying drone and ArduPilot that was published last week. |
I've just wanted to be sure that developers are aware of this issue. |
Yes VIO is quite sensible to vibration , Since then , I mount my systems on Damping Reduction Platforms ( 2$ @ Banggood), here is the T265 mounted on such a Damper Platform: |
@MartyG-RealSense video is from Thien Nguyen a PHD Student , I am mentoring on GSOC ( Goodle Summer Of Code) project this summer. You can watch his progress here: I asked him to upload on the RealSense forum once completed (2-3 weeks), we will certainly get in touch with you as you are the champion of the forum ;-) |
@patrickpoirier51 can you pls share what have changed after you used damping reducing platform? Is t265 performing better now? |
I haven't done comparison yet on T265, but on the setup showned above , that was the only way to make it work (as you can see the EKF was drifting until loosing track) Vibration tests is part of the experimentation that Thien Nguyen will be doing this week. He has a fixed mounted T265 on his setup and he will be inducing various degree of vibrations using an unbalanced propeller. We will upload Lab report here once completed. |
@urock, @patrickpoirier51, all our internal testing has been on Falcon 8s, and I haven't seen any issues with vibration. I can imagine that smaller drones might have different vibration characteristics. You should be able to filter the motor frequencies with a custom vibration insulator. |
Here is the Lab Report The T265 is fixed mounted to a 3D printed bracket, and take note that on the vibration test video, hoangthien94 had to manually hold and shake the Quadcopter in order to get the system diverge (degrade) for a short period. |
Hi @patrickpoirier51 , I am trying to use the T265 on a UAV for autonomous flight in outdoor environments. I have performed successful indoor autonomous flight, but as soon as I go outside the T265 pose estimate diverges and the drone loses control. I am guessing it is a mixture of my vibration and the increased scale of outdoor environments, since it works extremely well inside with the same vibration levels. I will be trying different vibration damping techniques soon to see if I can get stable flight. Do you have an idea of the maximum vibration that the T265 can operate in? In other posts you have said you've obtained accurate scale up to 60 ft with the T265. I am curious about your mounting configuration - was it forward or downward facing? Due to size constraints, I have to mount the T265 on the side of the drone at a 45 degree angle. Do you think that mounting configuration could be causing some of my issues? I am confident my pose transform is correct as I have successfully flown indoors. I have attached a picture of the drone and vibration on the T265 in m/s^2 in the UAV body coordinate system (x = forward, y = right, z = down). |
@alexbarnett12 I recommend you get a short angled USB3 cable because most of vibrations should comme from cable loop that catches airflow |
@alexbarnett12, I would concentrate on mechanical dampening. The T265 gets scale from two places, one is stereo and the other is the accelerometer. When you are 60ft up in a drone, stereo is close to useless, as the baseline is too small compared to that distance, so we have to rely on the accelerometer. I see acceleration noise greater than 1g! That's going to cause problems for scale recovery. |
Hi guys, let me put mine dime into the discussion. We have performed tests of T265 on vibration stand going though frequencies and acceleration amplitude in [2 - 150 Hz] and [0.5 - 8 g] measuring point when T265 would start loosing position. Results are the following You can see that with increasing frequency acceleration needed for T265 to loose position decreases from 8g to 1g saturating at 50 Hz. Saturation occurs because T265 can't report acceleration faster than 50-60 Hz. But this graf definitely shows that higher vibration frequencies have stronger impact on T265. Also I can report that we are flying with T265 without any dampening pads, it is firmly fixed on drone frame. But we have put a lot of efforts to flight vibrations on the frame itself. We are using best carbon props, motor mounts are firmly fixed, frame has to be solid without any loose parts. |
@alexbarnett12 would you be able to share your pose transformation for the 45 degree angled mount? |
Update from my last post: I was able to achieve stable outdoor flight by switching the Realsense mounting configuration to downward or forward facing. I did not have to make any changes to the level of vibration reaching the Realsense. I find it quite interesting that the 45-degree mounting configuration worked well indoors but not outdoors - I don't have the expertise to explain this. It might be something that the Intel Realsense team should look into. However, vibration is sometimes still a significant problem. Every once in a while, the Realsense will experience tracking failure upon takeoff outside. It seems completely random as I will perform a great test flight, restart the UAV, and then it will crash. I think this is probably because of my vibration levels, but with my size drone and my limited time and manufacturing capabilities, I cannot feasibly reduce the frequency or magnitude of the vibration. Here is a video of an autonomous flight test I performed where the UAV (slowly) navigates towards a yellow tarp. I never got around to testing at very high altitudes before COVID-19, so I don't know if the Realsense would experience tracking failure at 60+ feet. @MathiasThor I have included my 45-degree pose transformation if you would still like it. I can't recommend mounting it this way if that's what you were planning! :)
|
Thank you for highlighting the issues with pose under vibrations on T265. We have moved our focus to our next generation of products and consequently, we will not be addressing this issue in the T265. |
Issue Description
Hi everybody!
I have noticed when using T265 under vibrations above certain threshold it tends to loose its position very fast. Even if you shake camera in hand, not to mention using it on drone as I am doing.
This video also shows this issue https://www.youtube.com/watch?v=GhHvuAoFC6I
I hope firmware developers could do smth with it, right now it is impossible to use T265 under vibrations.
Meanwhile I will try to use vibration vibration damping pads.
The text was updated successfully, but these errors were encountered: