Thread: programming deadreckoning gps software?

Originally Posted by wi77iam

I'm liking the idea of just using independent sources of speed and heading, because that's really all you need to dead reckon, rather than some complicated accelerometer-corrected last known speed/direction. So that compass module would supply the heading (wish it was a phidget though). And for speed (and ultimate coolness), how about a Doppler speed sensor.

Are you sure that doesn't just measure the speed something is moving away from it? Besides, we could always hook up our own wheel RPM meter to the axle, right?

Oh, yeah, mount it at an angle. Duh. I was only thinking of pointing it straight down and why that wouldn't work

Wow thats expensive

Ok heres what I wanna try in the future, I think its should be quite accurate and cheap :

Take the speed by counting pulses generated as the wheel turn or by tapping into the speedometer. This can be auto calibrated by comparing the calculated speed with the GPS reported speed. Of course it can also be calculated by knowing how many pulses generated in one wheel revolution per sec and its circumference.

The front wheels angle can be measured by attaching a potentiometer (a bit tricky) or by counting pulses as the steering wheel is turned, a strip of equally spaced black lines on white is wrapped around the steering shaft. The black/white lines are sensed by an optical sensors...similiar to the way the mouse work (been mentioned above before).

Regarding the steering not linear, how about calibrating it? The car is driven around with for hours until the software can build up a complete look up table. The amount of calculated steering and the gps reported location are always compared until full calibration is completed.

The driving around calibration can of course also calibrate the speed. It can even be on auto calibrate all the time, calibrating the steering as well as the speed...atleast a messed up suspension and wearing out tyres are taken into consideration as it happen.

Driving the car hard, wheel spinning, handbrake turn, power slide...etc will of course mess up the auto calibration

Any feedback? doable? or just plain stupid?

Autocalibration just needs to take those things into account, and not calibrate

Originally Posted by GWRedDragon

Autocalibration just needs to take those things into account, and not calibrate

And how would the software know that the car is sliding, wheel spining? etc so that it wont recalibrate with those values?

Sound all too easy huh?

Another thought :

Most GPS receiver update the location once every seconds. If we can use the car speed and steering wheel postion why not fill in the gap in between those 1 second GPS updates? This way we can increase the overall position update rate by 5 per second lets say so then we can have a smother map scroll/update, very useful in a roundabout.

Filling in the gap from the last known location should be accurate since we only need to rely on predicting the car locations for a second.

Any feedback? been done before?

Originally Posted by Ricky327

And how would the software know that the car is sliding, wheel spining? etc so that it wont recalibrate with those values?

Sound all too easy huh?

Sure. Have it calibrate based on long term data, and filter out crazy spikes.