GPS - Is it all its cracked up to be?

GPS has become somewhat of a buzz word in karting data analysis since AIM introduced it on the Mychron 4. Many have waxed lyrical about its application and usefulness to determine driver performance (braking, lines, corner speed) but how accurate is it really;

Generally speaking GPS devices are accurate to within 15m (just look at your phone’s GPS). Some devices also incorporate WAAS (Wide Area Augmentation System) which can get the accuracy down to 3m, but still not good enough for karting. Accuracy can be further increased by using GPS/INS.

GPS/INS is the use of GPS signals and inertial navigation systems (INS). INS is the combination of accelerometers, gyros etc. If it is being used as an integrated INS/GPS, which allows the INS to interpolate the data between GPS samples, producing a higher data rate than what the GPS receiver can do on its own, and the GPS is used to null the errors in the INS then it is possible to increase the accuracy. These systems are used on aircraft, missiles and increasingly in some automotive test applications to determine roll, slip etc.

However this really only gets the accuracy down to 2m, still not good enough for karting.

VBox Automotive, a data acquisition manufacturer for the automotive industry claim much higher accuracy;

"GPS positional accuracy is subject to ionospheric interference, which causes a satellite signal to change in length as it travels towards the Earth’s surface. This means that, depending on the quality of the receiver, accuracy can vary by several metres.

VBOX data loggers are fitted with superior GPS engines that give better positional accuracy than those found in mobile phones and satellite navigation devices. By using additional signals other than those from the standard US GPS satellites, this can be enhanced further. The free SBAS services augment the positional capabilities of a survey-grade receiver to as good as 1m without the use of a Base Station.

However, by adding another GPS engine at a fixed position, the majority of errors can be calculated and removed from the final result. This is the function of a Base Station: because it remains stationary, it is able to broadcast correction messages to a roving VBOX, resulting in much improved positional accuracy.

By incorporating signals from the Russian GLONASS satellite constellation and employing a Real-Time Kinematic (RTK) algorithm, the VBOX 3i Dual Antenna RTK with suitable Base Station achieves a positional accuracy of +/-2cm."

I happened to use one of these systems at university, and sure they are accurate, but they were also £25,000. The price may have come down but they are still, I would assume, outside of the realms of most karters, plus they wouldn’t fit on a kart, the two GPS receivers need to be 2.5m from each other.

So the question is, have we been duped? Are we accepting lower levels of data fidelity for a buzz word? Why do the Tonykart factory team use a none GPS system? Or should we all go buy an ebox and all the extended sensors. Maybe AIM could tell us?

My personal opinion, GPS data is good but only when backed up by a suite of physical sensors and when combined with the correct physical sensors one can set up the kart without talking to the driver :wink:

3 Likes

Nice post.

I agree. I believe the Mychron4 and Mychron 5 augment the data using accelerometers with the 5 using a higher hz rate and glonass too.

My understanding is the addition of glonass does not increase accuracy more than the accuracy of the individual glonass or gps signal. However it does allow a quicker pickup and increased accuracy in “urban canyons” just because you have more satellites to choose from.

I wish AIM would just tell us how accurate it is, rather then having to guess.

1 Like

Actually, Race Studio does tell you how accurate your GPS positioning is in the output data. There is a channel called GPS Position Accuracy, which tells you how accurate your GPS data is down to the meter, based on the number of satellites that it picks up.

Ironically, I just finished recording an AIM video on this exact subject, which will be coming out in a few weeks.

1 Like

I stand corrected.

Although 0.5m is still a lot in karting. You couldn’t compare lines for example with any confidence.

I guess if the relative positions (between laps/runs) are consistent vs actual then that would be fine. But I can’t say of that’s the case or not.

Gps is accurate to within .3m at worst in my testing when using the new GPS08 module. I run concurrent ebox data with a GPS08 and the data matches up exactly. Sometimes speed is a little off because of the method used to calculate speed while turning but more or less it’s incredibly accurate. I also have spec’d out kits using top of the line MoTeC components where we’ve used a 6-axis IMU and 10hz GPS/GLONASS module.

There are far better ways to compare lines than the GPS track map such as gyro which is accurate to within a tenth of a degree on the GPS05 module. It’s probably better on the GPS08.

For what it’s worth the GPS is accurate enough that I can tell (through a few math channels) what the go kart is doing at any point in the corner in a way that’s been verified accurate by some of the best drivers in the sport. Dont over think sensors. AIM does a pretty good job sorting them out. :wink:

1 Like

Yep, I know that, obviously you know that. But most people who buy a GPS enabled datalogger don’t. My point of the original post is its probably not worth the expense.

I think I’m right in remember that with the mychron 4 there only GPS Gyro and Acc but with the mychron 5 there is a gyro and accelerometers in the head unit too.

I had quick scan of some data from Sonoma and its consistently 1.2 to 1.4m off (if the AIM pos accuracy channel is to be believed). The problem then becomes that to do any meaningful chassis testing that error is calculated through out.

Personally, I don’t use the GPS function to compare driving lines. I use it so that I’m sure to have a beacon that works, regardless of what track I’m at.

I’ll use a camera or, when I can find someone reliable enough, have a spotter with me at the track to help me determine driving lines. I can rely on the lateral and long g channels, combined with speed and RPM to help me figure out corners where I need to be more consistent.

That’s pretty neat. Gyro or Accelerometer? I’ve always found the latter to be very noisy in a kart, sometimes from mounting, but also due to the abrupt nature of the vehicle itself. Do you perform any smoothing on the channel?

GPS08 makes a HUGE difference. Original gps module is only GPS not GPS+GLONASS so it is over a meter off which is effectively useless for chassis tuning. The GPS08 is under a half meter every time I’ve used it. The problem I’ve found is if the MyChron 5 is mounted on the steering wheel, you actually have enough error in steering to throw more sensitive math off. But it does have an integrated IMU which gives you hardware accelerometers and gyros. Which is pretty cool but once you filter them down to an actually readable point it isn’t much more accurate than the GPS. Also, again, mounted on the rotating point of the steering wheel. I will often use a 3 filter if need be but most of the math channel stuff I do doesn’t usually require any.

A quick way to see how much more accurate the gyro is than the calculated in the module lateral acceleration is to try and build a track map based on 4 wheel vs 2 wheel (latAcc vs gyro). You can calculate latAcc from tangential speed and yaw rate and see how far off it is as well (it’s a lot). I guess my point is you can use gyro and especially gyro derived math and still be pretty accurate for line on track because you’re not relying on the GPS position sensor but you’re relying in the compass baked in which is substantially more accurate.

2 Likes

Exactly.

My concern is that your average karter pays for a GPS capable datalogger when there is little use for it and the IMU is mounted in a way that it isn’t much better.

It ends up being a waste money when AIM Drack Gold from the 90s would do a better job.

The GPS Accuracy AIM Learn Fast Video is live!

1 Like

I’ve found the GPS/GNS in the MyChron 5 to be just fine. I don’t use it to “compare lines”, because you can’t and shouldn’t try to with a $500 data logger. Get the GoPro out for that.

I find that my positional accuracy is usually pretty consistent when I’m comparing one track on one day. It varies track to track a little bit.

The biggest benefit GPS gives is being able to identify areas where deltas are occurring. Is GPS data perfect? No, but your average customer isn’t going to know what to do with a raw accelerometer channel containing lots of noise.

I reckon the new generation of GPS are pretty good. However, I still prefer to run a front wheel-speed sensor to back up any findings that the GPS tells me.

OR vice versa, if the wheel-speed sensor tells me that the driver found time by braking later for a corner, then I use the GPS to corroborate that data.

The Mychron 5 and Unigo 6005 have made significant improvements with the GPS, but if you don’t fully trust GPS (like me) then the Unigo does allow you to ignore GPS data, and use only the wheel-speed as the distance measure. For me this is a huge deal. You can switch between using GPS and wheel-speed for distance in the software and decide whether you believe the stories you are being told.

I’ve never been able to stop AiM data using GPS as its distance measure, other than by unplugging the GPS - but you can’t do that with the mychrom 5.

Using GPS to compare lines is a dangerous game for sure - but I have worked with a driver experimenting with wet lines, and the GPS did match exactly what he was doing on track, so we could measure whether riding a kerb was more effective than taking a wide line.

Are you correcting your front wheelspeed for left vs right turns, and for corner radius?

Interesting question. How is that done (Math channel) and what kind of % difference from F/L have you observed.

No, I’m not that technically astute, but I’m not looking for absolute numbers - I’m usually concerned with comparing what is going on at one particular piece of track, on different laps, or different sessions. Then I don’t need to control for those variables.

But usually I’m only using the wheel-speed occasionally to back up an argument, or throw doubt on evidence that the GPS has raised. So, for example, the GPS might show that a driver braked later and after the exit of the corner came out with a 2 tenth advantage compared to all the other laps in a session.

In that situation I’d look at wheel-speed data to see if it agrees that the driver did brake later. I’d also pay attention to when the last magnetic strip was hit, because that would help determine how reliable the wheel-speed data is

Need a math channel to calculate corner radius, then a second math channel to calculate the difference from where the measured tire is to the vehicle centerline.

The second math channel will have an if statement in there to account for left and right turns.

If you’ve got a wheelspeed sensor on the kart now, check out the difference between GPS speed and wheel speed as the kart is in a corner.

1 Like

I don’t personally use this function. But does the math channel calculating the distance from tire to center line account for the change in distance with the steering geometry. The distance from the pivot point stays the same, but the distance to the centerline decreases with lock and not in a simple arc.

All seems a little complicated when you can get the corner radius from;

radius = abs(GPS_Speed * GPS_Speed / GPS_LatAcc) * 0.02

0.02 is if you use GPS_Speed in mph, it gets you to radius in meters.