Understeer/Oversteer without steering angle sensor? Possible?

Submitted by tednet on Tue, 11/25/2014 - 14:03

Forums:

Hello!

I am using my XLoBorg for a car tracking computer using a raspberry pi. I want to have the ability to get some telemetry data. I do real-time calculations and log any dangerous driving (i.e. hard braking/accelerating).

I was wondering, is there a way to log if there is understeer or oversteer without having any knowledge of the steering angle? I have a GPS connected as well, but if it's possible just with the XLoBorg, it would be great.

Best Regards,

Teddy

tednet

Tue, 11/25/2014 - 14:33

Actually, I think i might have found a solution, depending on how fast my GPS can update.
We can compare the heading(yaw) of the GPS and the XLoBorg and depending on the difference we either have oversteer or understeer. You can read more about it here: http://www.aim-sportline.com/download/doc/eng/on-track_sessions/gps_mxl-...

However, if anyone has a solution that uses just the XLoBorg, I would love to hear it.

piborg's picture

piborg

Tue, 11/25/2014 - 15:32

Great you might have found a solution! The only caveat is the link you sent requires an gyro which the XLoBorg does not have. However, the system also uses:
"GPS, 4 shock potentiometers, steering angle potentiometer, front and rear brake pressure potentiometer, longitudinal accelerometer, lateral accelerometer, gyro, 4 speed sensors..." and CAN data. This may even include the steering angle in CAN data!

piborg's picture

piborg

Tue, 11/25/2014 - 15:26

Hi,
I certainly don't want to say this is impossible, but I would say quite difficult.
If you are not looking for any closed loop control, so you wouldn't for example try to apply brakes when under-steering or light a warning light, but simply record that an event happened, then it might be possible. If you are after closed loop control, then I would think the problem requires calculating the intended arc versus the actual one, and would require a steering angle sensor (or one could possibly derive this from tyre rotation speeds inner and outer to get an arc)

So, to hypothesize a couple of scenarios:

1) A normal turn - (black on picture)
In a normal turn, the car is driving forwards, at some point (a) the driver turns the steering wheel to the right. The body slowly starts to roll, the tyres are gripping well, and a centripetal force is somewhat smoothly applied and can be sensed by the accelerometer. By point (b) the wheel is at some angle (say 10 degrees) and this remains somewhat constant whilst executing the turn. At the end of the turn (c), the wheel is straightened, the body roll returns to flat, and the force decays to (notional) nil (d). The body roll and centripetal force decays slowly and smoothly.

2) A turn with under-steer - (orange)
The car is driving forwards, the driver turns the steering wheel to the right (a). The body slowly starts to roll, the tyres are gripping well, and a centripetal force is somewhat smoothly applied and can be sensed by the accelerometer. At some point (mid a-b on picture), traction gives way and the vehicle follows a less tight arc. The point where the traction gave way, the centripetal force will be different, and there will most likely be a change in body roll (the vehicle roll may even oscillate). At the point where traction is regained, assuming the wheel is still at the original angle, there would be another event where the body roll and the centripetal force would vary quickly. At the end of the turn, the wheel is straightened, the body roll returns to flat, and the force decays to (notional) nil. The body roll and centripetal force decays slowly and smoothly.

3) An over-steer turn - (blue)
This would probably be similar to the above, but I think it would be likely that at the end of the over-steer, a violent regaining of traction would occur and this could be monitored in both body roll and instantaneous acceleration. I'd imagine this as a large spike somewhere later in a turn. (Spike between b and c).

Take this with a pinch of salt, It's my best guess as I've never analysed this problem!

tednet

Tue, 11/25/2014 - 21:17

Wow, thank you for answering so quickly and with such great detail. This is probably the best support I have ever received!

I am still very new to using any kind of sensor. Isn't the heading that I can get using the magnetometer using the implementation (https://www.piborg.org/comment/189#comment-189) kind of like yaw? Anyways, so far the GPS heading, which is in degrees, does not seem to be too accurate. So comparing headings will probably be mostly useless, especially with a refresh rate of 1hz for the GPS.

No, the XLoBorg won't be used to control the car. It is for pure data logging only. Since the data will be uploaded in real time, I find it useless to upload telemetry data and use up all the traffic, instead of just uploading an event with abnormal g-forces.

However, I have no way of getting steering angle. My only sensors are one XLoBorg and the GPS. As far as I understood, centripetal force can be calculated just by using the accelerometer readings? So for that I need to measure the radius from the XLoBorg to the center of the car? Is this the proper way to the calculations or am I going in a completely different direction? http://www.ehow.com/how_12093135_use-accelerometers-measure-angular-velo...

Thanks again for being so helpful!

Subscribe to Comments for "Understeer/Oversteer without steering angle sensor? Possible?"