- March 16, 2016 at 9:55 pm #4736
I installed the IMU on some sort of small car, that has the ability to recognize if it is moving by using different sensors, that were already included.
Using the IMU I wanted to add a new dimension to the self localisation progress. I am having trouble recognizing the Position by using the double integral over time of the acceleration.
My idea is so far, that if a movement is recognized, the last few (10, 20 or 100) data acceleration Reads of the X-Axis, are used (by the double integral of each looptime) to set the current Speed to an absolute value, until the other sensors detect that the car stopped.
This right now is only used to recognize a straight drive,
without curves or something.
Sadly the precision of the Sensor isnt accurate at all.
The car drove 2m, but sometimes the sensor measures that the car moved 5m, sometimes it only measures 1m.
Did anybody else try to use the acceleration sensors to get information about the current position of some sort of car?
Best Regards.March 17, 2016 at 3:55 pm #4737Mark WilliamsKeymaster
What is the math you are using in your calculations?
have you applied any filters? low pass, high pass, median ?
Mark --OzzMaker.com --March 17, 2016 at 7:31 pm #4739
I am calculating an Offset using a mean Filter of the last 500 values so that the drive of the IMU is shifted around Zero.
And now, if a movement is detected I just take the last (10,20,100) RAW values minus this offset value. Therefore I see the real difference in the acceleration.
For example, the RAW values are from 100 to 200. I meanfilter these over a long period and shift them so that I see values around -100 to 100. Now if an acceleration occurrs I see like 10 values that are around -100 to 100 and inbetween those like 2 or 3 values in the size of 600 to 900. Those i sum up and integrate.March 17, 2016 at 9:06 pm #4740
I need to add that I do transform the RAW values into the unit [m/s^2] before integrating.
- You must be logged in to reply to this topic.