GPS Has a HUGE Problem That We Can Fix (RTK GPS)
video description
Date: 2026-07-10
Related videos
Comments and reviews: 20
erroneum
Your explanation of how GPS works is wrong. If you had a properly configured atomic clock to measure time with, and it was a known offset from the ones aboard the GPS satellites, then you could do direct trilateration to get a lock, but no consumer grade gear has that. Instead, it takes the time received by each satellite and takes the difference between two of them to form a hyperbolic surface upon which you must be, then uses 4 or more pairs thereof to narrow the set of possible locations to a single small bubble somewhere in the vicinity of earth (after accounting for limited precision. This can work because the signals being received from the GPS satellites are CDMA encoded, so they're all broadcasting at the same time on the same frequency, then you use a bit of math to extract a specific one's signal from it, but you can use the same math with different parameters to extract the signals for every one you can see from the same signal you've got buffered.
Other than that, great video.
reply
Your explanation of how GPS works is wrong. If you had a properly configured atomic clock to measure time with, and it was a known offset from the ones aboard the GPS satellites, then you could do direct trilateration to get a lock, but no consumer grade gear has that. Instead, it takes the time received by each satellite and takes the difference between two of them to form a hyperbolic surface upon which you must be, then uses 4 or more pairs thereof to narrow the set of possible locations to a single small bubble somewhere in the vicinity of earth (after accounting for limited precision. This can work because the signals being received from the GPS satellites are CDMA encoded, so they're all broadcasting at the same time on the same frequency, then you use a bit of math to extract a specific one's signal from it, but you can use the same math with different parameters to extract the signals for every one you can see from the same signal you've got buffered.
Other than that, great video.
reply
castle-quassel
Nice video! The reason your robot struggled so much at the end is the classic Skid Steering Trap.
With a single RTK antenna, the heading is derived from the motion vector. When a tank does a point turn, its GPS position doesn't change, making its yaw mathematically undefined. Asking a magnetometer sitting on a metal chassis right next to high-current motors to guess North in that exact moment is pure masochism. ;)
The standard industry fix for UGVs is a Dual-Antenna GNSS setup (like a Unicore UM982 or dual F9P) with a 60cm baseline. I used this for my UGV. It gives you True GNSS Heading via carrier phase interferometry instant yaw, zero drift, and 100% immune to motor EMF. The IMU should only be the Kalman-filter backup for the three seconds the robot drives under a tree.
reply
Nice video! The reason your robot struggled so much at the end is the classic Skid Steering Trap.
With a single RTK antenna, the heading is derived from the motion vector. When a tank does a point turn, its GPS position doesn't change, making its yaw mathematically undefined. Asking a magnetometer sitting on a metal chassis right next to high-current motors to guess North in that exact moment is pure masochism. ;)
The standard industry fix for UGVs is a Dual-Antenna GNSS setup (like a Unicore UM982 or dual F9P) with a 60cm baseline. I used this for my UGV. It gives you True GNSS Heading via carrier phase interferometry instant yaw, zero drift, and 100% immune to motor EMF. The IMU should only be the Kalman-filter backup for the three seconds the robot drives under a tree.
reply
AlanTheBeast100
There is nothing at all wrong with GPS accuracy. It was never meant to be more accurate than 10m or so on the civilian signal. (Indeed in the beginning the civilian signal was purposely inaccurate to deny accuracy to the enemy - better than 100m could not be counted on. And at the time we found this to be great.
The addition of WAAS/EGNOS (SBAS) has made many receivers accurate to the 2. 5m level in the right conditions (need a good view of the sky - the further north you are, the more of the sky to the horizon you need - not available world wide either - lookup SBAS.
RTK is a local solution only - further away from the reference station means growing error (about 1mm per km - not bad at all.
reply
There is nothing at all wrong with GPS accuracy. It was never meant to be more accurate than 10m or so on the civilian signal. (Indeed in the beginning the civilian signal was purposely inaccurate to deny accuracy to the enemy - better than 100m could not be counted on. And at the time we found this to be great.
The addition of WAAS/EGNOS (SBAS) has made many receivers accurate to the 2. 5m level in the right conditions (need a good view of the sky - the further north you are, the more of the sky to the horizon you need - not available world wide either - lookup SBAS.
RTK is a local solution only - further away from the reference station means growing error (about 1mm per km - not bad at all.
reply
tomboxyz5564
Some great cheap GPS receivers for RTK GNSS are LC29H, they come in different wariants (LC29HEA for a 10Hz RTK rover, LC29HDA for 1Hz RTK, LC29HBS for a base station) those modules use the newer L5 band (along L1) instead of L1L2 in the ublox ones, I use a LC29HEA for agricultural precision navigation.
In most places you can use some form of free RTK service, there's community made ones, but also a bunch of goverment ones (like ASG-eupos in Poland) you might not get the best precision since you could end up 30-40km from a base station, but if you're trying to get started with RTK it removes one variable, get a rover working with a known-good base station, then you can make your own base station
reply
Some great cheap GPS receivers for RTK GNSS are LC29H, they come in different wariants (LC29HEA for a 10Hz RTK rover, LC29HDA for 1Hz RTK, LC29HBS for a base station) those modules use the newer L5 band (along L1) instead of L1L2 in the ublox ones, I use a LC29HEA for agricultural precision navigation.
In most places you can use some form of free RTK service, there's community made ones, but also a bunch of goverment ones (like ASG-eupos in Poland) you might not get the best precision since you could end up 30-40km from a base station, but if you're trying to get started with RTK it removes one variable, get a rover working with a known-good base station, then you can make your own base station
reply
elektron2kim666
I accidentality made my GPS projects able to connect to the laptop map software just by forwarding the GPS text to USB (ESP32. Mostly, I made clocks due to the problems explained in this video. I can think of a lot of ways to do this like a little matrix in memory and building a fake lawn in that and the triangulation from bat technology. Each beacon could be air plane tech and so on. My best idea is to start simple with a tiny boat and let it find home beacon 1 which is an infrared remote. It will circle when it can't find it. Then you point the remote at the boat and it will give it a go in that direction. In the back there could be out-mode to not wait too long.
reply
I accidentality made my GPS projects able to connect to the laptop map software just by forwarding the GPS text to USB (ESP32. Mostly, I made clocks due to the problems explained in this video. I can think of a lot of ways to do this like a little matrix in memory and building a fake lawn in that and the triangulation from bat technology. Each beacon could be air plane tech and so on. My best idea is to start simple with a tiny boat and let it find home beacon 1 which is an infrared remote. It will circle when it can't find it. Then you point the remote at the boat and it will give it a go in that direction. In the back there could be out-mode to not wait too long.
reply
MichaelWoodrum
This proves, in some way, that when I am stuck in a loop of thinking or that anything I eventually come up with, someone will push out an answer.
I have an rtk setup from 7 years ago. I used the base stations from my state to correct. Now I am building a zero turn mower automation after making it remote control.
I was stuck on a few things but location being the hardest to overcome without it being fuzzy.
So, true to the hacker ethos, the internet has provided. I know there is a rule but, I am old and annoyed.
Once again, you deliver with great content that then makes me spend more money.
reply
This proves, in some way, that when I am stuck in a loop of thinking or that anything I eventually come up with, someone will push out an answer.
I have an rtk setup from 7 years ago. I used the base stations from my state to correct. Now I am building a zero turn mower automation after making it remote control.
I was stuck on a few things but location being the hardest to overcome without it being fuzzy.
So, true to the hacker ethos, the internet has provided. I know there is a rule but, I am old and annoyed.
Once again, you deliver with great content that then makes me spend more money.
reply
vj68895
I'm wondering how long you have to run that base station so it knows it's own location accurately enough for surveying purposes. In surveying you normally set up the base station on known point for rtk-gps. But it's possible to set it up anywhere and wait long enough for the base station to solve it's own location globally accurate. This combo Scott used is good enough when you only need relative coordinates to the base station. But if you want exact location in national coordinate system or in globally accurate GPS coordinates then you must know more accurately the base station's location.
reply
I'm wondering how long you have to run that base station so it knows it's own location accurately enough for surveying purposes. In surveying you normally set up the base station on known point for rtk-gps. But it's possible to set it up anywhere and wait long enough for the base station to solve it's own location globally accurate. This combo Scott used is good enough when you only need relative coordinates to the base station. But if you want exact location in national coordinate system or in globally accurate GPS coordinates then you must know more accurately the base station's location.
reply
RCSIM-social
Awesome video, Scott! To drastically cut down the cost of RTK GPS, you can use the Waveshare LC29H GPS/RTK HAT (based on the Quectel LC29H chip) paired with a free or local NTRIP service. This completely bypasses the need to buy and set up a second physical base station. I use exactly this configuration on a Raspberry Pi for my autonomous RC car project, and it achieves centimeter-level accuracy for a fraction of the price of dual-band U-blox modules. Definitely worth checking out for a budget DIY project!
reply
Awesome video, Scott! To drastically cut down the cost of RTK GPS, you can use the Waveshare LC29H GPS/RTK HAT (based on the Quectel LC29H chip) paired with a free or local NTRIP service. This completely bypasses the need to buy and set up a second physical base station. I use exactly this configuration on a Raspberry Pi for my autonomous RC car project, and it achieves centimeter-level accuracy for a fraction of the price of dual-band U-blox modules. Definitely worth checking out for a budget DIY project!
reply
SquintyGears
From your description there doesn't seem to be anything special in the RTX kit that you couldn't do yourself with just having 2 of the basic gps kit.
Having to develop your own correction algorithm with good performance is a bit of a challenge and would probably lose you some precision. But the bought kit gets you bellow 1cm and a DIY version would be within 5cm. Not the kind of error margins that matter for whole house.
So what exactly is the secret sauce for RTK to cost 10x regular GPS modules
reply
From your description there doesn't seem to be anything special in the RTX kit that you couldn't do yourself with just having 2 of the basic gps kit.
Having to develop your own correction algorithm with good performance is a bit of a challenge and would probably lose you some precision. But the bought kit gets you bellow 1cm and a DIY version would be within 5cm. Not the kind of error margins that matter for whole house.
So what exactly is the secret sauce for RTK to cost 10x regular GPS modules
reply
CrispyCircuits
About ten years ago I was trying to visit Glacier National Park in Montana. Like an appropriate idiot, I chose the wrong Amtrak passenger train stop. So I had to hike and hike. Cell phone service in Montana is very bad. It's extremely rural there. But my phone could work with just the regular GPS service. I also had one of those water filters with garbage water in, clean water out.
I'm glad to see this nice GPS upgrade. But that's way too expensive for a cell phone or tablet.
reply
About ten years ago I was trying to visit Glacier National Park in Montana. Like an appropriate idiot, I chose the wrong Amtrak passenger train stop. So I had to hike and hike. Cell phone service in Montana is very bad. It's extremely rural there. But my phone could work with just the regular GPS service. I also had one of those water filters with garbage water in, clean water out.
I'm glad to see this nice GPS upgrade. But that's way too expensive for a cell phone or tablet.
reply
erikcederb
I am currently building some things with the UM982 module from Unicore communications, that is a dual-antenna, triple-band, all-constellation GNSS module which with RTK data fed over NTRIP gives me both centimeter level position and a true heading down to about 0. 1 degree accuracy with an 1. 8 meter baseline between the two antennas, in a reciever that costs under 200.
It also supports Gallileo HAS fully, so I can even get sub-meter precision without the RTK link here in the EU.
reply
I am currently building some things with the UM982 module from Unicore communications, that is a dual-antenna, triple-band, all-constellation GNSS module which with RTK data fed over NTRIP gives me both centimeter level position and a true heading down to about 0. 1 degree accuracy with an 1. 8 meter baseline between the two antennas, in a reciever that costs under 200.
It also supports Gallileo HAS fully, so I can even get sub-meter precision without the RTK link here in the EU.
reply
michaelsegel8758
cm level precision has been available for a very long time.
There are reasons why your cell phone is not that precise in terms of actual location. This is why you see A-GPS which not only uses GPS signals but also signals from base stations to get a more accurate position.
And then there's the issue of the accuracy of the map. 15yrs ago the map was only accurate to 1. 5m This was because your car is larger than 1. 5m
There's more, but you get the idea.
reply
cm level precision has been available for a very long time.
There are reasons why your cell phone is not that precise in terms of actual location. This is why you see A-GPS which not only uses GPS signals but also signals from base stations to get a more accurate position.
And then there's the issue of the accuracy of the map. 15yrs ago the map was only accurate to 1. 5m This was because your car is larger than 1. 5m
There's more, but you get the idea.
reply
dmangTV
well thats disappointing. was really hoping this was more of a howto with some cheaper boards and not buy this $600 kit. thankful for the comments here. knew about AGPS and DGPS but wanted to stick with real RTK for some drone/mower projects. the boards mentioned in the comments so far: ZED-F9P is $250, UM960 are $150, Quectel LC29HEA are $60, and you need 2 for base station and rover. if the Quectel can actually deliver real RTK then I'm sold
reply
well thats disappointing. was really hoping this was more of a howto with some cheaper boards and not buy this $600 kit. thankful for the comments here. knew about AGPS and DGPS but wanted to stick with real RTK for some drone/mower projects. the boards mentioned in the comments so far: ZED-F9P is $250, UM960 are $150, Quectel LC29HEA are $60, and you need 2 for base station and rover. if the Quectel can actually deliver real RTK then I'm sold
reply
nandoanalog
20 seconds into the video, I suspect he's going to talk about rtk. I got a couple rtk modules from aliexpress and on the first day I couldn't get rtk fix no matter how good the antennas placement and shielding was. Left the chips plugged in all night, and the next day I got sub cm accuracy, which felt strange, but was very real. rtk works weirdly well with proper antenna placement and clear sky.
reply
20 seconds into the video, I suspect he's going to talk about rtk. I got a couple rtk modules from aliexpress and on the first day I couldn't get rtk fix no matter how good the antennas placement and shielding was. Left the chips plugged in all night, and the next day I got sub cm accuracy, which felt strange, but was very real. rtk works weirdly well with proper antenna placement and clear sky.
reply
killersiggy
With the two F9Ps if you put them both on your machine you can have the heading be determined by the ublox module doing phase shift calculation removes imu and calibration from that. Basically set one up as moving base and have it output the rtcm into the other one. Might not get fixed but still gets to float and sub 10cm without external corrections. Heading is then in the UBX. NAV. PVT msg.
reply
With the two F9Ps if you put them both on your machine you can have the heading be determined by the ublox module doing phase shift calculation removes imu and calibration from that. Basically set one up as moving base and have it output the rtcm into the other one. Might not get fixed but still gets to float and sub 10cm without external corrections. Heading is then in the UBX. NAV. PVT msg.
reply
greatscott
About 25 years ago I worked at a company that installed a few stationary stations called Differential GPS for the military.
That station had a GPS receiver, and, knowing it's own real position, it would calculate the error on that same instant and transmit it to a large radius.
The mobile Differential GPS receivers that caught the signal could then correct their GPS receiver position.
reply
About 25 years ago I worked at a company that installed a few stationary stations called Differential GPS for the military.
That station had a GPS receiver, and, knowing it's own real position, it would calculate the error on that same instant and transmit it to a large radius.
The mobile Differential GPS receivers that caught the signal could then correct their GPS receiver position.
reply
PeriliousPaddy
GPS has always been a rough number, during peace it is accurate to within 30 meters, during war times it then gets narrowed down to 3 meters accuracy. I had a friend at University in 2008 that was an avid paraglider, he was always cursing GPS for bad positioning but loved paragliding during the Falklands war as his landing spots were a lot safer for him.
reply
GPS has always been a rough number, during peace it is accurate to within 30 meters, during war times it then gets narrowed down to 3 meters accuracy. I had a friend at University in 2008 that was an avid paraglider, he was always cursing GPS for bad positioning but loved paragliding during the Falklands war as his landing spots were a lot safer for him.
reply
Stef. Cata051
Would it be easier to use a local positioning system in order to align a car for charging or for a robot lawnmower. For the car example the hardware could be baked in the pad itself and it would only need to communicate with the vehicle when the vehicle is in that area but for the lawnmower it would be a little bit more difficult but no too dificult.
reply
Would it be easier to use a local positioning system in order to align a car for charging or for a robot lawnmower. For the car example the hardware could be baked in the pad itself and it would only need to communicate with the vehicle when the vehicle is in that area but for the lawnmower it would be a little bit more difficult but no too dificult.
reply
Ryan_Tidwell
Cool video. None of the projects you demonstrated needed a signal to go to space. A robot lawnmower for example.
I'd love to see a diy solution for a small area. Three or four tiny transmitters on the corners of your yard would be very similar but most of the problems with gps would not be an issue.
How accurate could you get it
reply
Cool video. None of the projects you demonstrated needed a signal to go to space. A robot lawnmower for example.
I'd love to see a diy solution for a small area. Three or four tiny transmitters on the corners of your yard would be very similar but most of the problems with gps would not be an issue.
How accurate could you get it
reply
xxportalxx.
Maybe I'm wrong, I haven't spent much time on the subject, but if you're resorting to base stations it seems to me that using gps like this is wicked overkill, and largely just throwing money at the problem. Seems there should be a cheap and clever way to get equal or better results using base stations for positioning than this.
reply
Maybe I'm wrong, I haven't spent much time on the subject, but if you're resorting to base stations it seems to me that using gps like this is wicked overkill, and largely just throwing money at the problem. Seems there should be a cheap and clever way to get equal or better results using base stations for positioning than this.
reply
Add a review, comment
Other channel videos
