Autocross Setup W/ GPS and GIS - Guaging Interest

Hello, my first post here.

I have been developing procedures to utilize consumer grade GPS and free/open source software for autocross setup for my SCCA region.  Long story short, each individual cone can be dropped on the course with an accuracy of about 3'.   I plan on refining the procedures for my region, regardless of interest here.

If there is enough interest, I'd like to document the process and procedures for use by others.  The software is rather cumbersome to set up initially, especially for those without a fair amount of GIS and CAD experience.  Documentation aimed at providing simple step-by-step instructions would be a large undertaking on my part. 

Once GIS is set up for a given venue, what I'm developing can provide the following with a reasonably small effort:

• Simplified collaboration and review of course design among designers, safety stewards, and event chairs.

• Reasonably accurate development of venue site drawings, including lot boundary, striping, curbs, light poles, large vaults, fences, gates and site access for export to CAD.

• Import of CAD course design files, in a real-world coordinate system, to utilize GPS capabilities.

• GPS based course setup, eliminating dead guesswork and significantly reducing the need for measuring wheels, tape or pacing.

• GPS based “as-built” of a course after setup, capturing changes made during course refinements.

• GPS based survey of venue site conditions, including peaks and valleys, poor pavement areas, small vaults and other features to consider in course design.

• Ability to export course design to Google Earth (KMZ files), for easy distribution of layouts to participants.

• Archiving course designs for future reference.

• Preliminary assessment of potential venue locations.

• Ability to create various site maps and exhibits without significant drafting effort.

This is pure tech geekery - who's geek enough to be interested?

I might be, but it's a bit of a stretch

Just an idea, you might be better off using a highly accurate local positioning system for recoding the course layouts and GPS for positioning the entire course, since the relative distances between the cones are more important than where the whole mess of cones is. Maybe something like this:

https://www.pozyx.io/

In reply to GameboyRMH :

That's really cool!  Back 9 years ago, when I was trying to make a DIY racing data acq- that would have been a very handy thing for me to get.  Back then, I was going to stitch together the fastest GPS that I could get with accelerometers.  But my passion died, and I never had to go into that really tough math to connect it together.

That box takes care of all of that.  Which would be incredibly handy for autocrossing.

GameboyRMH said:

I might be, but it's a bit of a stretch

Just an idea, you might be better off using a highly accurate local positioning system for recoding the course layouts and GPS for positioning the entire course, since the relative distances between the cones are more important than where the whole mess of cones is. Maybe something like this:

https://www.pozyx.io/

Interesting.  Unfortunately I'm a Civil Engineer, not a programmer.  Getting coordinate correct location data to interface with GIS and CAD software is the key to most of what I'm doing.  I'll look at it some more, but getting something like this to work is certainly way beyond my depth.

I think it will be of limited use with only 3 feet of accuracy, when I course design, I radically change speeds and approaches through elements with moving cones less distance than that. 

(comment meant as constructively as possible)

In reply to Apexcarver :

Yes - understood.  I do the same and make adjustments for speed and flow after the initial lay-down of cones.

What's lead me down this path is the lot we use has no stripes, expansion joints or easy landmarks to help transfer a design to the ground.  It's just a lot of black asphalt.  Courses end up being a very vague copy of the design by eyeballing element locations, or we'd need to spend hours with a tape and wheel.

I tried my GPS routine at our last event, and laydown of cones was surprisingly fast and accurate.  This left lots of time to refine the course once it was initially on the ground.

I've found that setup crews, myself included, want to get in and out as fast as possible.  Spending 2-3 hours attempting to get it close to the designed geometry wasn't leaving much time and patience to hang around and do the refinements.

You don't have GPS-enabled cones that are self-aware and report their position to Skynet automatically?  Amateur.   

In reply to chad1376 :

So would you pre-design a course on paper, and use the tool to put the cones down, or design the course on the fly, and then locate the cones that you can publish a course map?

For me, it would have been the latter- I never once put a course on paper, and I thought our courses were pretty fun.  But other than a conceptual map, people had to really rely on the course walk.  Having something where I can GPS each cone, and then publish it somehow would be pretty cool.  (that being said, I would not be a customer either way, it's been 9 years since I designed a course).

In reply to alfadriver :

We've had some very good courses by just throwing cones out without a plan, but also many abominations, arguments, and strife.   Personally, I do much better with a plan beforehand.  I hate moving whole sections of a course 5-times.  I also think it gives the course designer more control of getting things done.  The no-plan method leaves too much leeway for all the setup crew to stand around and argue over what they want to see (and move whole sections of the course around 5-times.)

There's still opinions on what could be done better during setup, and many times the ideas are better.  With a plan, though, I say "lets set it up as designed, then see how it drives."  We make changes after it's substantially complete and the course can be seen as a whole.  Not just one element at a time.

Have you looked into using a more accurate GPS receiver, like a Trimble R1? That'll get your cones located to a few inches. Are you using mobile devices to collect and/or display your spatial data?

I'm a GIS analyst who hangs out here so this sounds pretty rad. There's at least one more of us here on GRM too. 

In reply to thatsnowinnebago :

Ohhh, that looks sweet!  But for $2,500, no.  I'm trying to do this in a way that any club can put this together for next to nothing.  I know our club starts getting uncomfortable if we need to buy more batteries for the timing equipment ;) 

Of course that wouldn't prevent someone else from buying a better GPS and getting those accuracies.

Cell phone and super cheap GPS isn't good enough, with accuracies on the order of 30.'  WAAS enabled GPS is needed  It's ghetto differential GPS developed for aviation.  A WASS USB receiver can be had for $30. 

I bought one with a 5-hz acquisition rate for $90.  It's more responsive as you are walking around on site.

Location from the receiver is displayed directly in the GIS software, on top of the mapping and course layout (I'm using QGIS - it's free and does what I need it to do.)   You need to carry a laptop around, with the receiver to get it all to work.  An old used convertible tablet (just a Core 2 Duo) with pen input is working good so far.

Ha, yeah, those receivers aren't cheap. 

Can QGIS do point averaging? I know Esri's Collector app can. That takes longer to get your data but it gives your higher accuracy. I've only worked in Esri shops so I've never gotten a chance to mess around with the open-source stuff. 

In reply to pinchvalve :

You jest,  but rc and robotics equipment is suprisingly cheap. A gps-aware cone dropping robot is totally a thing one could make. 

Edit: Spoke too soon. https://www.newcivilengineer.com/latest/costain-robocones-could-save-lives/10035596.article

In reply to thatsnowinnebago :

I don't know.  I'll take a look to see.

I think, on a practical note, you wouldn't want to hang around in one spot too long waiting for averaging, since there are so many cones to lay down.  It might be good if you wanted better location data for recording site issues (vaults, dips, poles etc.) though.

To be honest, I'm not an expert in any of the deeper functions that might be available.  I'm learning much of this as I go.  Maybe that'll be a good thing if I try to write something up, since dumb mistakes and the learning process are still fresh in my head.

I do somehow want to more rigorously test the accuracy of the GPS data, with my own equipment, to get a better feel for how big the error might be.

Nugi said:

In reply to pinchvalve :

You jest,  but rc and robotics equipment is suprisingly cheap. A gps-aware cone dropping robot is totally a thing one could make. 

 

Edit: Spoke too soon. https://www.newcivilengineer.com/latest/costain-robocones-could-save-lives/10035596.article

Holy hell - forget what I'm doing.  Just have the cones set themselves up! 

In reply to chad1376 :

I've got a decent methodology for testing accuracy between two devices if you want to hear it. 

In reply to thatsnowinnebago :

Sorry for the slow reply - trying not to kill to much real work time on the 'net.;)

Hit me with what you got!

My biggest hurdle now is finding decent free and capable CAD software.  I've tried about everything I can find.  Most do about 80% of what I want, but lack a few important features to bring it all together.  I might need to explore some of the lower-end CAD packages ($100 range) to see if they are more useable.

This assumes you have access to survey-grade equipment at work. 

Mark a grid of like 6 points, more than 6 ft apart (the goal is to make sure the points don't fall within the expected error of each other). Then capture each point with your survey-grade setup and export the co-ords to a CSV. Collect the same points with your autocross setup. Import both sets of point into your GIS program and run the QGIS equivalent of Esri's Near tool. That'll give you a table of each pair of points and the distance between them.

Or you can use a known location, like a survey benchmark or control points, and collect that location with the autocross setup. Then compare the results as above. 

I like it.

My office doesn't have survey (we sub it out), but there are BLM benchmarks in the desert near my house.  The desert is good, so I don't get run over standing around and marking things in the street.

I found NGD database site that has locations and data control sheets for each.

I'll bet if I use some Google-foo, I can find a program or site that will process the data if QGS doesn't.  If anything, I can just visually see how much the GPS track wanders as I stay in one spot.

Now to find the time to actually do it.

Babbling  update:

I did some testing to see how much the GPS location wanders over time.  Both are about 1/2-hour of collecting location data.

The first is with a Globalsat 363-S4 (5hz) reciever, running on Linux.  Getting Linux to consistently assign the same USB port and link to the driver was a headache.  As you can see, for some reason it's not utilizing WAAS differential gps.  Error is about 2.5 meters.  Not good.  This is disappointing, since I understood the GPS unit does the differential correction internally.  Something about Linux and the driver is making this problematic.  I'm done with fighting Linux for this application.

This is the same hardware, with Windows 10 and the factory driver.  Ahh, much better.  Error is 1 meter.  Importantly, the location "walks" across this pattern, it doesn't randomly jump around.  This means that cones set up at about the same time will have a much smaller (relative) error - say on the order of 1' or less.  The fuzzy car you  can see in the aerial is a Golf R for a sense of scale.

I have another el-cheapo tablet (from eBay) and more gps units that should be here by the end of the week.  Our next event is this Sunday, and I hope to have two GPS computers for course setup on Saturday.  My goal is to lay down the initial course in 1/2-hour (more time for flow refinements.)