HKUST-Aerial-Robotics / VINS-Fusion

An optimization-based multi-sensor state estimator
GNU General Public License v3.0
3.45k stars 1.37k forks source link

Is this not a good suggestion to use imu of realsense d435i on smart car? #191

Open ghhhgfgh opened 2 years ago

ghhhgfgh commented 2 years ago

I have tried MONO+IMU, and stereo + imu using D435i in VINS-fusion, but both cannot get a good result.

I have used Kalibr to calibrate the extrinsic parameters between IMU-mono/stereo and the imu noise and bias random walk. Also, I used the OpenCV to calibrate the intrinsic parameters of the mono-RGB/stereo camera.

Some details I have also noticed such as closing the IR projector of the D435i camera.

The first time I think I cannot get a good result may be about the time synchronization Td of camera and IMU so I use the Td calibrated by Kalibr as initial and open the estimating Td function but it didn't have a better result.

So it only makes me wonder if the IMU data is inaccurate or not... In my smart car, the vibration sources are the LiDAR RPLIDAR A2 and the fan of the onboard computer (But I think it's not the main problem that makes the imu data not good?) Also, the wheel motor uses the hub motor and doesn't install damping, so when moving on the ground, there is some vibration(Maybe it's the main reason?).

So, I wonder to know is there anyone using d435i on the car who can get a good result of VINS? Or it's not a good suggestion to use imu of realsense d435i on smart car.