Thread: programming deadreckoning gps software?

DOH! You're right. If you dropped the sensor, it would read 0g, not 1g. Crud.

OK, so we should be able to correct for tilt if we tap into the OBD. Maybe.

Dominik, where did you get that compass Phidget? I don't see it on their website. I believe compasses (magnetometers) have the same problem with tilt, unless you have a 3-axis one where you can correct for it... Tell us more!

Is this the compass? http://www.robot-electronics.co.uk/s...CMPS032004.htm

Originally Posted by jpelzer

I give up.

Maybe a frame of reference based discussion might help? Ignoring all other forces, gravity will cause an object to accelerate (roughly) towards the core of the earth at 9.8m/s/s. In the case of our sensor, let's assume it's affixed to our car. That means that the car now exherts an equal and opposite force on the sensor -- no matter if it's on level ground or driving up a hill. If the car is not accelerating, neither is the compass. The compass will always measure the same acceleration as the car because their frames of reference remain congruent.

Now if accuracy and GPS plotting is the goal here; driving up a hill may still introduce a error in your plotting. The problem is that your speed measurement correspond to a distance vector in three dimensions and your lat/lon positional data (relative to the earths surface) is only two dimensional. In other words, your map distance traveled will be (measured speed * interval * cos (angle of incline)) assuming a constant speed and angle of incline. For long highway trips up/down steep grades this will probably introduce substantial error. Unless you have three dimensional mapping data and/or some additional sensors, I'm not sure there is away around this.

UPDATE: I just did some reading on how accelerometers work and apparently they ignore gravitational forces. Hence an accelerometer sitting on a table will measure an acceleration of 1G strait up.
http://www.lunar.org/docs/LUNARclips...erometers.html

I guess this means we need to distinguish between what an accelerometer reads and actual acceleration.

Originally Posted by jpelzer

Is this the compass? http://www.robot-electronics.co.uk/s...CMPS032004.htm

I have that one and some code I wrote to read it over the parallel port. I haven't tried it in my car yet though I'm bit concerned that it might not be too accurate because of all the metal and other magnetic interference in the car. Walk up to your car with a normal compass and check our what happens. Maybe that's why dead reckoning is usually done with gyros?

Originally Posted by jpelzer

You're confusing yourself. An accelerometer measures acceleration. Period. Not gravity. If we wanted to measure gravity, we could use a scale, and tilting it compared to the earth would change the reading.

If you have an accelerometer, and it is parallel to the ground, not moving, it will measure 0g. If it is 45 degrees to the ground, not moving, it will measure 0g. 90 degrees to the ground, not moving, 0g. It won't measure any value other than zero if its actual speed is not changing.
.

Think of the accelerometer as a small mass on a spring oriented along an axis. Acceleration along the axis will make the mass move against the spring and this movement can be measured as a voltage. If you place an acc (I'll abbreviate since I'm tired of writing out accelerometer) oriented on the long axis of the car, you will measure acceleration forward and back. (In an aircraft this is called thrust.) An acc oriented perpendicular to this and parallel to the ground will measure acceleration left and right (yaw). And another acc oriented perpendicular to both of these will measure acceleration up and down (pitch). An acc on the fourth axis (roll), would measure acceleration of a car on a banked turn.

The pitch acc (and possibly the roll acc) will also sense the effects of gravity, which is indeed an acceleration. A scale measures weight, not gravity.

The thing to remember is that these accelerations being sensed are only used to apply corrections to the current direction and speed of travel of the vehicle. So with the last known speed (from the GPS before it loses lock), the acc will sense acceleration or deceleration of a certain amount, that can be used to correct the last known speed, which can then be used to calculate the change in latitude and longitude. Similarly for change in direction, which is measured as acceleration in the left/right axis, and applied to the last known heading. Again used along with the speed to calculate new latitude/longitude.

The pitch acc would sense the car going up or down a hill, but I don't think this would be important to the lat/long calculations because the thrust acc will always be providing the corrections to the speed, which along with the direction sensor is all you need to update your position.

This is the essence of dead reckoning navigation.

Originally Posted by rando

I have that one and some code I wrote to read it over the parallel port. I haven't tried it in my car yet though I'm bit concerned that it might not be too accurate because of all the metal and other magnetic interference in the car. Walk up to your car with a normal compass and check our what happens. Maybe that's why dead reckoning is usually done with gyros?

Don't confuse dead reckoning with inertial navigation, which does use gyros.

Yah, what he said. It all makes sense now.