Based on the data we've been collecting the acceleration values contain biases, see the attached picture with the histograms.
Specific values for our device are xbias = 30, ybias = -3, zbias = -83
Can you confirm that this is the expected behavior?
Are these bias device specific, i.e. if I collect data with a different device, the biases will be different?
Thank you
PS. It looks like I cannot attach my png file. What it shows is the histograms peaked at the values xbias = 30, ybias = -3, zbias = -83
Acceleration
Re: Acceleration
Bias? Do you mean defaults? The accelerometer will read whatever is acting upon it. By default, gravity will always pull 1G towards whatever is "down". For instance, if your OBD port is pointed down, you should have a full bias on one of the 6 axis'. But almost no OBD2 port is pointed straight down.
-
davidmakovoz
- Posts: 8
- Joined: Tue Feb 07, 2017 1:57 am
Re: Acceleration
@protomor, thank you, it makes perfect sense.
The thing is, I am a data scientist, I just got the data. The actual device is half-way across the world, I only saw it on the computer screen
How stable the device is in the port, does it sway in any direction?
The thing is, I am a data scientist, I just got the data. The actual device is half-way across the world, I only saw it on the computer screen
How stable the device is in the port, does it sway in any direction?
-
davidmakovoz
- Posts: 8
- Joined: Tue Feb 07, 2017 1:57 am
Re: Acceleration
I've been thinking how to go about recovering the actual acceleration from the OBD data.
In general, there are four unknowns: 3 Eulers angles and the magnitude of the acceleration.
Since I only have 3 measures, a_x, a_y, a_z, there is no way to recover all four unknowns.
One way to solve it would to conduct a calibration drive to measure the yaw. Yaw is not going to change, unless the car starts spinning, which hopefully won't be happening too often.
I'm new to this area, I wonder if somebody could provide guidance in terms of what the standard practices in this area are.
In general, there are four unknowns: 3 Eulers angles and the magnitude of the acceleration.
Since I only have 3 measures, a_x, a_y, a_z, there is no way to recover all four unknowns.
One way to solve it would to conduct a calibration drive to measure the yaw. Yaw is not going to change, unless the car starts spinning, which hopefully won't be happening too often
. I'm new to this area, I wonder if somebody could provide guidance in terms of what the standard practices in this area are.
Re: Acceleration
I can't tell you about someone's OBD port. I think they should be fixed and don't move. But some people do stupid things to their cars.
You could also use an extension cable and mount the unit somewhere. The 6 axis accelerometer is in the plug itself and not read from the OBD port. So you can mount it anywhere in the car. If you're trying to derive facts about the car, you might want 2 accelerometers. One in the middle of the front axle and one in the middle of the rear axle. But that's physically impossible.
You could also use an extension cable and mount the unit somewhere. The 6 axis accelerometer is in the plug itself and not read from the OBD port. So you can mount it anywhere in the car. If you're trying to derive facts about the car, you might want 2 accelerometers. One in the middle of the front axle and one in the middle of the rear axle. But that's physically impossible.