IMU Fun at 1 to 3am

October 1st, 2006

So I decided to try using the IMU to give me my displacement as I move around one night around 1am. I knew the IMU had some error when it was just sitting on the table but I didn't know exactly how much. Also I wanted an excuse to ride the ENS elevator up and down... In conclusion, our IMU has way too much error after integrating the acceleration values twice. Reseting my counters and riding up to the 5th floor I'd get a value of +1000 (not a clue as to the units but that wasn't important yet) Reset again and ride back to the lobby gave -1500. Standing still the value drifted slowly down and sped up as the first integrator grew.

More fun? I'm only testing in the Z axis so I go back the the trailer and set the IMU on the table. A big problem is Z axis always is about 0.988 but varies a bunch like +- .05 at least. I tried some filtering but that was quickly a dead end if the speed you accelerate does not exactly match the decceleration rate. At least that what seems to make sense since the ways I filtered was to not change the output number unless it was greater than a certain threshold.

The night wasn't a total waste since I did end up making a USB to DC plug cable (so I could power the IMU off the laptop) and I got to ride the elevator alot (despite it being pointless since the stationary test failed.)

Conclusion

for those who don't want to read the above, the IMU won't work by itself to provide accurate positional displacement...

Darn.