This week, my focus was on implementing the accelerometer portion of the Kalman filter and debugging issues related to bias drift. Right now, I’m working on getting the Kalman filter to report a steady absence of movement when the IMU sits still on my desk. I’m getting close but I still have some drift issues to work out and I believe that some adjustments with the Kalman filter error matrices should help. I’m currently driving the accelerometer Z channel to 0 and averaging the values of the first 300 samples (6 seconds) of the IMU lying still. After I’m satisfied with the level of drift present in the still readings, I’ll move to testing with data sets that contain 2D movement. Then, once Luca finishes with the gyroscope Kalman filter, we’ll integrate the two to have 3D calibration and thus 3D positioning. The trick will be to eliminate the ~1g of gravity that will dispersed throughout the various axis when the IMU isn’t lying on a flat plane.
I participated in several meetings this week with Luca where we discussed the schedule moving forward and the current bias drift issues we’re having. Professor Mukherjee provided some good guidance in our group meeting on Friday and if I’m not able to make significant progress on the drift issues through the weekend, I’ll go back to him for more help.
Below is an image showing the current drift that we’re trying to eliminate in the readings for the IMU standing still.
0 Comments