GPS instability

[ClydeCrashKop]

Just out of curiosity, I tested this GPS Module with Ceramic Antenna for DIY Handheld Positioning System for $7.86
https://www.dx.com/p/open-smart-gps...ino-apm2-5-flight-control-480918#.WiDKElWnH4Y because without moving, the Latitude & Longitude numbers go up and down independently. It seems like it is wandering around but stays within a 25 feet diameter circle for the most part. I guess that is pretty good trying to triangulate my location from 3 or more moving objects in space.
That is not good enough for a GPS controlled lawn mower.
I took a few sets of data with a Stationary GPS sampling once per second then plotted the 3 least significant digits in Excel.

 

[alec_t]

Informative. Don't land-based positioning systems, for e.g. farming and surveying, use 'Differential GPS', involving a fixed local beacon as well as the satellites, to get extra precision?

 

[ClydeCrashKop]

I should look into that. I guess you get what you pay for.
These guys claim centimeter accuracy https://www.swiftnav.com/piksi-multi
and it is good enough for autonomous vehicles • Starling Positioning Engine
And even cloud based • Skylark Cloud Corrections Service
I was fascinated with LORAN when I had it on a boat. That was all land based timing triangulation. I don't remember how accurate it was. I knew of a boat that set his autopilot way point on a buoy. He hit it dead center!

 

[Dr_Doggy]

and a common problem with GPS,
alot of the apps use algorithms to compensate .... i only know cos if i go off route with my phone it takes about 5 seconds to realize it and re coordinate

until the point where you go off route far enough the gps will assume its just drift and keep you moving on the original course at your calculated speed

I hope you get this working it is definitely a problem that you need a dog fence to run one of these mowers ....

OMG ima genius!!! ... .I just had a thought.... how hard would it be to build our own local triangulation system ... so like we set up 3 beacons for reference in the corners of the yard ... then the bot would have better resolution ... also there are less clouds throwing things off down here

... i am thinking more now ... a wifi module has rssi for wifi signal strength ... i wonder if we could use that for local pos (we dont need gps coorinates really just local boundary coordinates) with 3 routers in the area(dont even need to be connected) we could maybe do a pinpoint with that?

 

[dr pepper]

LoranC is (was) accurate to about 50 meters.
It has been put back into service here in the Uk, but is unuseable for navigation, just for frequency.
Rssi on wifi isnt a great deal of use for navigation directly, if you walk round a room it will go all over the place, not necessarily in relation to the distance to the router.
A dog collar type boundary system would be effective, I think they use Rf around 100kc.

 

[Dr_Doggy]

thats what they use now is the dog collar system .... it seems crule to me
more importantly the cost to install a dog fence is more expensive than a state of the art mesh network

 

[rjenkinsgb]

Signal strength alone is no good for distance or position over a small area. Phasing effects and reflections can cause the signal to vary dramatically within a couple of feet at VHF and above.
That's why such as WiFi gear has multiple antennas, so not all combinations of path distances are generally in nulls at the same instant.

Anyone that's worked with radio comms is familiar with mobile flutter and even aircraft flutter (the reflection from an aircraft overhead causing dropouts).

A high precision system needs either distance by signal timing from each beacon, or be able to measure bearing to beacons.

It does not help that precise distance by radio timing is inherently difficult as you need nanosecond accuracy just to get to about a one foot error margin; fractional nanoseconds to get better.


For a short range system around a house for such as a mower, an ultrasonic beacon system could possibly work?
That makes distance by timing easy for a simple MCU, it's about one millisecond per foot for sound.

 

[Dr_Doggy]

but what about in a mostly open yard, there will be less reflection there, a quick calculation for std wifi range is about 150ft and signal drops out at about -70 & - 80, so thats about 1db per half foot(ok so i round off non linear ), I have noticed when scanning inside homes for wifi that a stationary item will fluctuate about 2db....

plus not sure of the scope for others but i am calculating for a urban yard where most yards are less than 1/4 acre ... and there is a wifi broadcast in every direction (in this situation you wouldn't need the "satellites" since the robot would load the map based on the SSID's near by) ... but even for the larger rural home without wifi you could set up something like this around the perimeter:

take one of these:
https://www.aliexpress.com/item/ESP...VSVuGCh0j3QAJEAYYBiABEgI1WfD_BwE&gclsrc=aw.ds
and put it in one of these:
**broken link removed**

.... and maybe add a RF trigger for the bot to wake it from sleep

and there we go for only 30$ we have our satellite station
.... put about 10 or 20 around the yard (every 50ft) (still less than fence) ... and now we have multiple points to navigate from even if one or 2 satellites drop out(neighbours move...ect) ... since we dont even need network access to measure the SSID and RSSI

would that be able to compensate for Phasing effects... or maybe 2 radios on the bot itself?

 

[dr pepper]

Give it a try, though you might be dissapointed.
Imagine 1000s of rubber balls being dropped from a great height into said garden, that is what microwave Rf would do, bounce all over the place.
Waves hitting the ant from different directions at the same time add & cancel each other confusing the issue.
Wifi only has a length of a few cm, to build a reasonably simple radio locating device a wavelength longer than the accuracy you want would be in order.
You can get into detecting unreflected wavefronts on rf signals, that gets real complex real quick though, or if you could somehow get your wifi router to 'ping', that too would probably be a tricky session with the router setup.

 

[Dr_Doggy]

OK the results are in !!!!
I am only pushing this cos i really would like to see a resolution to this problem.

So i wrote an app to pole the RSSI from 3 extenders at the same time ... running unpaired to any network. I would like to mention that due to the limitation of the android SDK my samples were taken at every 5 seconds ... however I think a faster polling from arduino would yield the same results.....

1)The first test was a linear test on a single extender done on a concrete floor with lots of objects all around to cast noise and reflections .... the phone was in a stationary position for the polling.

2)The second test underneath the first one was also a linear test on a single extender but elevated about 3ft from the floor ... notice at the 3ft mark was a coat rack which caused an offset at that point. also there may be bouncing from my unsteady hand holding it.

3) the following 4 tests to the right of the first two is a 20 sample triangulation between 3 extenders with a stationary device at 4 different points. The map illustrates the positions and distances involved.

The forced columns are just random numbers I wrote in.
The rar file is a spreadsheet with the results.
THe red are the extenders .
the green are testpoints
the blue are tv
the yellow is me!
tests were conducted at the same height within themselves

SO..... still think it would not work well enough??

 

[Pommie]

I read about a system in an aircraft that used GPS to indicate roll and yaw. The two GPS units were at the ends of the main wing. The actual readings weren't important just the difference in the readings. It was claimed to be very accurate. I'm guessing that they both had the same "wandering" error so it cancelled out. I think it would work equally well for a lawn mower. You need to be careful that the GPS units don't average their position and just give raw coordinates.

Mike.

 

[ClydeCrashKop]

SO..... still think it would not work well enough??

That might work if you map it to the lawn and average it. The third column of data that I couldn't use would have stabilized it a lot.

The two GPS units were at the ends of the main wing. The actual readings weren't important just the difference in the readings.

I like that! For $7.86 I ordered another one. If they wander in the same pattern it should be easy to calculate the position of the mobile unit with reference to the stationary one.

 

[ClydeCrashKop]

Anyone that's worked with radio comms is familiar with mobile flutter and even aircraft flutter (the reflection from an aircraft overhead causing dropouts).

I live about 15 miles from an airport but sometimes I am in the flight path. An airplane may have caused this issue.

 

[ClydeCrashKop]

The two GPS units were at the ends of the main wing. The actual readings weren't important just the difference in the readings.

I like that! For $7.86 I ordered another one. If they wander in the same pattern it should be easy to calculate the position of the mobile unit with reference to the stationary one.

Dual GPS readings wandered independently by 0.00420 minutes or more of longitude, which I calculate to about 25 feet.
That was fun setting up one GPS, Arduino, wireless transmitter combo to send position info to a GPS, Arduino, wireless receiver combo that sent both GPS positions to the PC.

Its not close enough to work with so I won't bother plotting the difference.

 

[dknguyen]

Are you going to try dead reckoning with wheels odometry + GPS next? Seems that would stop the GPS from running wild. How badly do lawmower idler wheels slip?

 

[ClydeCrashKop]

I think that is beyond me right now. I have rotary encoders, ultrasonic, accelerometer, magnetometer and other sensors. You never know. Seeds planted in my head tend to grow. Maybe I will watch for a better GPS. It looks like GNSS is better but way expensive.

Meanwhile, I have motorized drive wheels, a WiFi camera and an RC airplane controller with receiver that has 2 channels of PWM. So I will play with the mechanical end of it that I have been talking about in this thread. Back EMF issues

 

[Pommie]

Have a read up on "Cosine direction matrix".

Mike.