@LuckyBird wrote a very good post integrating ardupilot and Intel realsense T265 camera. However with T265 discontinued, we need other VIO solution. The goal of this project is to build a small and low cost VIO system with common hardware.
This project is based on VINS-Mono, it is a real-time SLAM framework for Monocular Visual-Inertial Systems created by HKUST Aerial Robotics Group. Thanks a lot for their great work.
Hello, I’ve reviewed the Ardupilot code modifications you made, where you added the ‘highres_imu’ message to retrieve IMU data from the FCU. I’m curious, why didn’t you utilize the existing ‘raw_imu’ message? The only apparent difference seems to be the timestamp and accelerometer units.
Why did you mention that the raw_imu accelerometer data is truncated? It’s simply divided by GRAVITY_MSS, converting it to units of ‘g.’ Afterward, the Raspberry Pi can perform the necessary transformations to obtain the original IMU data.
The IMU signal need to be generated from the FC as it provides minimal latency and correct usec timing (because of the real time OS ChiBios) thus allowing the correct pre-Integration and propagation of the IMU state within the estimator.
You may certainly work with Raw IMU and process within Raspian if you would like to compare accuracy and consistency of VINS.
Do you mind sharing a power train details? I’m considering reproducing this with quite similar hardware, possibly designing a custom frame for the setup.
I would also like to try using RaspberryPi 5 with two cameras (since it has two camera MIPI ports now) and vins-stereo. Before Pi5, this was only possible using a special board multiplexing two cameras or using a compute module.