Hi
I have IMU sensor that gives me the raw data such as orientation, Angular and Linear acceleration. Im using ROS and doing some Gazebo UUV simulation. Furthermore, I want to get linear velocity from the raw IMU data. If I do integration over time there will be accumulated error and will not be accurate with the time when for example the robot makes turns.
So If I use
acceleration_x = (msg->linear_acceleration.x + 9.81 * sin(pitch)) * cos(pitch); acceleration_y = (msg->linear_acceleration.y - 9.81 * sin(roll)) * cos(roll);
So integrating linear acceleration is very bad,
Velocity_x= Velocity_old_x+acceleration_x*dt;
because integrates the acceleration without taking into account any possible rotation of the sensor, which means that the results will probably be terrible if the sensor rotates at all.
So I need some ROS package that takes into account all this transformation and gives me the most accurate estimation of the linear velocity. Any Help? Thanks
Dear Bojan Andonovski,
I suggest you to see links on topic.
https://www.youtube.com/watch?v=nQZeAafDEJE
https://www.youtube.com/watch?v=hN8dL55rP5I
http://docs.ros.org/en/melodic/api/robot_localization/html/preparing_sensor_data.html
https://upcommons.upc.edu/handle/2117/119640?show=full
http://wiki.ros.org/ingeniarius/ForteRC
http://wiki.ros.org/common_msgs/1.0_Migration_Guide
https://groups.google.com/g/hbrobotics/c/aNGzfOp_jZI?pli=1
Best regards
The problem I face to solve is the integration of the 3D acceleration data to estimate the Velocity will accumulate a big error over the time. The aforementioned integration typically leads to significant error accumulation. So im looking for some robust integration that can minimize that error.. So optimization,, or something. .
There is method proposed by Sagawa for correcting such errors. So Im looking for some better methods and possible C++ /ROS implementation or package.
Dear Bojan,
what do you mean with "rotating sensor"? The sensor should be fixed regarding position and orientation. Otherwise you will never get rid of that error - even worse: you might not know the size of the error produced, as it might be at random. However, it the sensor's "rotation" is caused by rare events, you could try to recognize such events and re-initialize/re-calibrate your position-sensing device afterwards. If it occurs more frequently, re-initialization/re-calibration on a regular basis might help.
Generally, as long as you use integration of differential sensor data, your position will always be biased by an increasing error. As already suggested by Mohamed-Mourad, you need additional sensor data and perform sensor fusion to minimize the error. You can also constantly estimate the current error. This does not eliminate but reduce the problem to some degree. Alternatively you can try to re-initialize your position data at some known positions (using recognition of some kind of beacons with known positions - just like in classic naval navigation).
Sensor fusion might even help to detect events causing your sensor to "rotate".
Sensor fusion is the second step i will perform after estimation of the Linear Velocity. So the first error i got is regarding that integration the acceleration. So my goal is to find some way to overcome (as much as possible) that naive integration and minimize the error( as much as possible, and yes I know that there will be always some ). Is it clear now?
Yepp, I think got it.
You could try applying a Kalman filter (https://en.wikipedia.org/wiki/Kalman_filter) to your problem. It is widely adopted for inertial short-time navigation. However, If you want to continue this type of positioning for a longer period of time, increasing positioning errors are inevitable.
How to apply KF to the integration part? So, simple KF will reduce the error of the integration over time?
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science