Earlier this year we reached out on ROS Discourse for sample datasets as we worked on an online utility for processing IMU data. I’m happy to say the service is now live! And that early feedback lead us to launch with support for processing rosbag files.
https://imuengine.io allows you to easily process IMU data (e.g. accels, gyros) into motion outputs like slope, tilt, or 3D orientation. You can upload data, run it through sensor-fusion engines, and download results all from your browser.
I know being a paid service isn’t ideal for personal projects or students. If this fits your needs but cost is the barrier then please do DM me. A fair number of folks use ROS for work and hopefully this can be useful to them. Feedback from all is appreciated. I’m part of the small founding team and we are busy adapting the service to fit the needs out there. Below is a bit more detail on why we launched.
Why We Launched imuengine.io
Prior to this, measuring motion was possible by either (A) buying a custom sensor or (B) logging raw data and using an in-house sensor fusion library.
Option (A) works great if you need real-time outputs and have a sizable sensor budget. There are great specialized sensors for construction, marine, energy, medical or robotic applications. But you pay with time to understand how to use the sensor and keeping sensor firmware up to date is non-trivial. (Case in point: the $25k of sensors ONAV bought in 2017 and 2020 run firmware that is now 6 and 3 years old, respectively.) When the sensor has left the manufacturer facility, its performance will likely never get better.
Option (B) is buying general purpose sensing hardware (e.g. imu chip or package) and customizing a sensor fusion software package to use it. The result is a much cheaper BOM, and by leveraging an open source or commercial software library, sensor fusion can be done in-house. An initial effort will go into sensor selection, calibration, and packaging. And a larger time investment will go into writing/choosing a sensor fusion library and tuning it to the application. Without on-going maintenance, the sensor fusion software can suffer the same fate as Option (A) firmware and risks becoming a liability.
With imuengione.io we manage the sensor fusion software and make it easily usable by any application which logs imu data and has internet. Think Google Docs for motion sensors. Come back from the weekend and the accuracy will have improved and additional features added. We want engineers using the service to deliver results faster and enable lower-cost sensors to really punch above their weight.
Here’s the direct link to try it out: imuengine.io/engine/