Question for machine control guys

Start thinking a 3d and curved planet.

Most (all?) coordinate systems are planar

Over a large site, earth curvature is significant. Grid to ground separation is significant especially if you are far from your coordinate system's central meridian.

A site calibration creates a plane that fits to an average over your on-site control - and matches the gps to that - but it is not going to be perfect. And if you are keen, it will also try to model the geoid to elipsoid varition in your area.

But here is the kicker - most engineers think and design only planar, and most cant actually think in 3d.

It is absolutely vital that the original control survey, the design and the machine control are all in the same terms - and that's why, yes you should have a site calibration, it should be the same one used for the original control and for the design.

I know of one job that was so large they needed six separate calibrations, depending on where they were. They allocated machines to specific areas and specific base stations and had to be really really careful to ensure that the right ones were used, especially when moving machinery from area to area.

I've been told that the machine control has to calibrate. I don't understand why that would be.

I've also been told that there can only be one calibration and that they will only do it once.

My recommendation was that no calibration be applied.

The site has been mapped in two stages using the same projection parameters. The first mapping was done in 2004 using NAD83(Epoch1993) state plane and Geoid03. The second mapping was tacked onto the original mapping and was finished this summer combined both projects cover almost 200 sq. miles and one control network.

Since construction will take place in numerous sites in the area which will be tied together the machine control guy wants only one system and they want a calibration.

I just don't see a reason to calibrate. I've never gotten good values doing that and more problems are created than anything that is solved. There is over 4' in changes in geoid heights across the site and I don't understand why any calibration is necessary.

I've worked in the entire area without one and don't understand why they can't, although I've never set up machine control receivers.

I'm with you on that one MightyMoe, I fully agree with your statements. We finally had an opportunity to work with machine grade gps last year. After talking to the machine grade gps surveyor I was very interested in the face that they calibrated to my control which was already on the State Plane coordinate system. They said that they had to...which I don't understand because we survey all over the area without a calibration just fine.

Of course this site was not setup for machine control so the calibration did not fully encompass the site which led to much confusion, but that's another story altogether.

Yeah, I don't get it. Sounds like they are going to do one of those set-ups where the base is put on a point and they locate a couple of points with RTK then do a calibration; and that will be fixed as the calibration from then on!!!!

And I gave them all the L,L,H N,E,El information which they could use to calibrate with without even leaving the office.

But that would only establish state plane info horizontally and a plane vertically which is wrong because there is so much change in geoid heights in the area. Seems like a real mess being created.

There's some weird thing about calibration that has cascaded down from the early days of RTK.

I've got a stupid question, since I haven't worked with machine control.

It would seem that the design would be done in NAD83 (probably via SPC) and NAVD88 and must be ultimately be converted to XYZ or some other ellipsoid based coordinate system, since that is what GPS is using.

Does this happen in the office, or in the field units, and does the conversion use the best geoid model? With a significant change in geoid height, that would seem important.

That's why I'm confused. The control is all SPC and Geoid03. Control was all done with static and adjusted. The XYZ is just a projection of the L,L,H information. So why calibrate? Why introduce error that doesn't need to be introduced. This isn't flat country and the Geoid slopes are complicated. There is no plane that will do a good job simulating elevations in the area; Geoid09 or 12 would be better but there was already so much mapping and design already finished using Geoid03 that everyone considered it the best solution.

I don't know what software they are using, but I use Pocket 3-D on my Hiper +'s. Pocket 3D forces us to 1) have the base on a fixed (control) point, then we must 2) localize on 1 or more additional control points (for rtk). I have found that if you have a project that is big, as in long or wide, the more control you have around the site the less error you will have.

Plain and simple, us dirt guys have software that works for us, some is limited and we can't do 'in the office' things that you can with your survey software.

I am not sure how a job is initialized on a dozer or grader however.

In my case, the digital site file(s) are taken in Carlson Takeoff, after selecting boundries, existing topos, design topos, building pads, pipe, and what ever else you want. Some time these polylines on the drawing have to be identified as what they are.

What I am getting at here is the Orig plot plan with the existing contours was done normally in SPC, so the design engineer will build on that plane. Doing this keeps everything (especially elevations) correct.

Thing is, you sometimes have to figure out what the surveyor did a couple of years ago for some developer etc, and get back to the plane he used.

You calibrate to orientate the base to the site control.

I use between five and seven calibration points (nothing more than control points with a x, y, and z).

And I also make sure that these points are past the extents of the site, so that everything that I stake is well within the calibration points.

It's a real problem for this project, the more I talked to the machine guy (who really understands what he's doing) the more I realized there is a good chance for a train wreck.

The site covers about 200 Sections of land. The first location covers about 8 Sections in the very NE corner. The equipment needs to have a calibration file imputed to make it operate. To change out this file in every machine would be a major undertaking. However, to extend this first projection across the entire 200 Sections will be a disaster.

Why it isn't possible to just enter the projection parameters into the file is something I can't get my mind around. I don't know if this is just old equipment, old software or just how the machine control industry works.

Back in the mid nineties I used to calibrate jobs, then I wised up and found better ways to work. Now it is rare that I will do it and only under duress.

Granted I only setup "pad sites" (Shopping Centers, Colleges, etc.), but this is all you are doing, orienting the model (linework, surface, points, alignments) to the points in the field. It is not uncommon for me to have files that are in a local datum and not in SPC. If your existing conditions plan/site survey says control point 1 is at x Northing and y Easting with an elevation of z, you have to calibrate (which is slightly different depending on the software/equipment, we only use TopCon & Trimble) to the field location of that point (which the software then compares 5000,5000,500 to the lat,long,elev).

Every project is different, but once observations on each point have been taken (30-60 seconds is typically considered enough due to the little change in the location of the satellites within a 5 minute time window), you will have to choose the points that are to used as your "God Points" (common term in this area...but the points that measure best upon themselves, while encompassing the entire site).

This means that the control file now changes the alignment horizontally of the model to be rotated slightly to fit the field points in the dirt and then vertically your model is "shaped" to hold the point elevations you see fit.

TopCon is a little easier to setup the control file, the Trimble you must use two files from the site controller to create one file to be used in the equipment.

This is the easy part...building the models is not...

I can see the need for "calibration" if the design was done, say, from total station measurements and no good GPS data.

But if the control was established by long averaging sessions of GPS, then those points have very good XYZ numbers and the model is built on them. Probably by converting XYZ to SPC and elevation. Why can't the resulting design be turned back into XYZ in the office and given to the field to be used verbatim? I'd assume the RTK base is on one of the control points used in the design.

That would let the office apply whatever geoid it wants in converting back to XYZ. The field equipment wouldn't need to know anything about planes, scale factors, rotations, or geoids.

After all, the control may be more accurate than the day-to-day repeatability of the RTK. Why does it need to be stretched or rotated to fit whatever errors the RTK gives you? Don't the RTK positions have zero bias on average?

I'm just guessing here, and would like to get educated. So far most of the discussion has been operational with no specifics as to what conversions are done where.

You have to understand what is going on in the equipment control box to get an idea for the reason for calibrating a site. Aside from getting a simple 3d location, the equipment is also trying to control numerous valves and sensors instantaneously to get the blade into the correct location, up and down, side to side, all while updating while the machine is moving. I think it's probably a lot simpler for the systems to work with a single correction from a calibrated base rather than all the extra stuff that would have to happen using SPC and a geoid.

I'd think the simple solution for a project like that would be a simple calibration at each site. If I'm reading your post right, every site would have its own model anyway, not like everything would be in the same one, that would be a disaster. Each calibration needs at least 5 control points surrounding the site. If it' not there, you need to add it.

That's been my observation.

To the extent of my knowledge, this is done because the software is built as a "one size fits all" approach. Just because Joe Blow Surveying says their points are so accurate, are they? Is there something wrong with their equipment that they don't know about? Did they perform all their collection in a manner that guarantees absolutely no error? I know I'm going to check all the points that will dictate this $$$$$$ job and not assume everything is correct.

The systems that I am familiar with are Base/Rover RTK, not network. These systems don't "know" anything about SPC or Geoids, strictly numbers in, numbers out. So simple factors of scale and rotation can be applied once to the model and be done. Why build so much extra into a system that is making thousands of calcs per second, while controlling a piece of equipment that is moving?! It is truly amazing how so much technology is applied to move some dirt.

Now having used TopCon much more than Trimble, I know for a fact I can put in not only the coords (typical), but the lats and longs as well if someone were to provide these. And using both of those pieces of information I could create a control file (.gc3) without leaving my desk or getting my boots dirty. Using just that information is a whole other discussion...

What TopCon does is compare your coords (say CP1 is N5000,E5000) to your lat long, then compares both of these to CP2 (N5100,E5000). So the difference of your lats and longs should show CP2 100.00' north of CP1, not 99.96' or 100.04'. Now move onto CP3 (N5000,E5100), does this measure due east of CP1 based on the lats and longs recorded (and 100.00')? Any difference between these points will cause the rotation or scale factor.

But the difference one would see between RTK and static, should not be "intolerable". If it is, well the dozers, graders and shovels are not going to be that great investment they thought they were going to be... if my RTK reading is 0.04' feet off for staking out sidewalks, who says the guy pulling his rusty mangled tape off my offset is putting it in the right spot?

So very many variables to be considered when having an electronic file move dirt...

Thats all I know...

> I'd think the simple solution for a project like that would be a simple calibration at each site. If I'm reading your post right, every site would have its own model anyway, not like everything would be in the same one, that would be a disaster. Each calibration needs at least 5 control points surrounding the site. If it' not there, you need to add it.
>
> That's been my observation.

:good:

I do not understand their problem with changing sites in the machines. If they feel their operators are not intelligent enough to navigate the files, they could always leave the cf cards at each site instead of in the machines.

If a job is on State Plane, then I export a DC file and then either use Site Vision office to convert it to a .cfg file for the machines or send it to the guy that does the models and he does the same thing using TBC. No calibration needed.

If it's a job using State Plane with a scale factor applied, it doesn't work so well. Exporting the site from Site Vision Office to google earth will show the site in the wrong place by the same amount of the coordinate shift from the scale factor. It does the same thing on the machines. It's typically a 1000 foot shift in my area.

For these jobs I prefer to do an "office calibration". Create a box with with four points that encloses the jobsite and enter those points by Lat Long Height into the data collector. I add a few points for a check, a point in the middle of a small site, extra points on the long side of a rectangle, several points along a long road, ect. Export those same points by North East Elev. Then import those points back into the data collector with a different point name. Typically I use G100, G101... for the geodetic points and C100, C101... for the corresponding coordinate points. Then I do a calibration using G100-C100, G101-C101,...

The Machines do not like large files with a geoid model in them. For these I take the geoid out of the calibrated file and use the extra points to see how the elevations compare. After calibrating just inverse between G100 and C100, the same with the rest of the calibration points. If the change is more that we can tolerate then we break the file down into smaller segments.

You can also get error if you are in a State that uses a Lambert projection and the calibration creates a Mercator projection. I have fixed that in the past but I don't remember exactly how I did it right now.

Sorry about getting so long winded
James

Thanks, sounds like the machines want a "simple" calibration to work, which is what I'm being told. Not sure just how much error will be introduced by doing that over such a large area. They will not accept the idea of different calibrations for each site as equipment will move between all the time and there is work to do between the sites. They have done similar (although not quit so large) projects before but they were in flat land.

In some areas we work in locally there is .5' of geoid height change per mile and calibrations don't work. This area isn't that extreme but it will be a challenge to make it all work-about 4' in 15 miles east to west. They tend to extend their base out quite a bit farther than I do while controlling machines. So one base and repeaters of some kind will control a number of machines over a large area.

If its Trimble each calibration would be stored with each site. Seems it would be a simple matter for the operators to switch files when they switch sites. As for getting it on each machine, the sites with their calibrations are stored on a single thumb drive. They automatically load when plugged into the box on the machine. Same thumb drive would load as many boxes as needed.

No matter about the geoid. Use it to set the control points for each site, then just calibrate to those control points. I assume the control points all tie to each other. If not you have bigger problems.

For clarification are you suggesting they have a single model that covers 200 sq miles? And that they are covering all that with a single base?

For clarification are you suggesting they have a single model that covers 200 sq miles? And that they are covering all that with a single base?

Yes, the mapping is all using one projection and yes it covers the entire area. They will do design with the contours already generated. And quite a bit has already been done using that information. The question is how to make the machine control work.

I've been told that they will not use more than one calibration for the entire area; they don't want to create more than one site calibration, they want the machines to operate seamlessly between the areas. And I agree with that assessment. To create more than one calibration and work between the sites would cause havoc. As far as covering the entire area with one base; I don't know. Right now they have one base, but as they expand I doubt that will work, but there is plenty of control for anywhere they wish to go.