[chbot] Differential GPS

[Charles Manning]

On Wednesday 29 May 2013 13:53:48 Robin Gilks wrote:
> Greetings all
>
> I'm just mulling over ideas for a walnut picker-upper and one of the
> requirements will be accurate positioning to make sure that I don't run
> into trees or irrigation nozzles. I think that if I can get below 0.5m
> accuracy within 300m of a fixed reference then it would be doable.
>
> Does anyone have information on such a system, perhaps transmitting in the
> 433MHz ISM band at very low power.
>
> TAPR did have one but the Motorola Oncore VP OEM GPS receiver they used is
> no longer available so it looks like the whole project is dead.
>
> I'm sure with the cheap GPS receivers available now there must be
> something available that can lock to a defined set of birds so that mobile
> and static receivers look at the same ones and correction information can
> be sent at a low rate from the static to the mobile station.
>
> Thoughts appreciated...

I lived and breathed GPS when I worked for Trimble for 12 years. I spent a lot 
of my time working with agricultural guidance (tractor steering etc).

Rather than tell you how DGPS works, it is probably better to just let you 
read up on various web sites.

Perhaps you don't need DGPS at all. DGPS gives you absolute accuracy (ie. 
absolute positioning). All you typically need in agricultural settings is 
relative accuracy. We were able to get better than a foot withregular GPS. 
RTK gave us 20mm.

ie. Say your gizzmo always starts off at a fixed point, then what matters is 
the drift once you have set this going. We found that since Selective 
Availability was turned off, relative accuracy is often pretty good without 
any correction at all.

A lot of the accuracy you get will not just depend on using DGPS or not, but 
also on the type of receiver you are using and the filtering etc in it.

Trimble receivers have always been position centric. ie. they always try to 
give you the best position. That is not what you want for agricultural 
applications. For example if a SV falls out of solution or a new one pops up, 
then a position centric solution will take that into account and try to give 
you the best guess position even if that hops 5 metres. In most agricultural 
setting you don't want the position to move unless there was real movement. 

We (Trimble Ag group) fiddled with the filtering etc and were able to add 
various fudge factors and algorithmic tweaks etc to get that working better.

These issues are particularly important when you are working under and between 
trees because you will be losing/gaining SVs all the time and their signal 
strength varies a lot. Depending on the filtering  and the solution 
calculations used in the receiver, it might end up weaving along.

You would probably do well to take a GPS receiver and log positions while you 
walk straight lines between the trees and see what results you get. Run a few 
tests at least a few hours apart to test under different SV geometries.

-- Charles