14mm GPS accuracy on a HOMEBREW bot!

[Scott Monaghan]

Hi all,

My Twitter buddy Jason just posted this awesome article about a bot that uses RTK GPS to achieve 14mm GPS accuracy with ROS2.

Check out: https://medium.com/exploring-ros-robotics/how-to-use-rtk-gps-on-a-ros-robot-a51e9aa2f2ab

Scott Monaghan

413.244.9344

scott.m...@gmail.com

www.linkedin.com/in/scott-monaghan

Resume: https://tinyurl.com/yxhq5eul

[Gmail]

Nice!

Thomas

-

Want to learn more about ROBOTS?

https://www.youtube.com/user/thomasfromriverside

On Nov 15, 2023, at 2:49 PM, Scott Monaghan <scott.m...@gmail.com> wrote:



--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/CAMzUCWOpsHiWvMGukrcF7VO4MMsMAdCN6ipfuneHkExyWP0eAA%40mail.gmail.com.

[Dave Everett]

Very interesting, thanks for posting Scott.

Dave

--

[Chris Albertson]

This is why they stopped encrypting the GPS signal.   When GPS was new the US military used to encrypt the last few bits of the GPS data and without the key you’d only have an approximate location.   But then some smart guy said that if the GPS receiver were placed on a well-surveyed location it could know what those bits should be and could automatically tell anyone within radio range the correct value of the bits.  This made encryption pointless

RTK is taking this same approach to propagation through the air.   It is random and makes GPS less accurate.  But a GPS at a surveyed point can measure the error and tell others.

But really there are not a lot of uses for ultra-accurate GPS.  Most of the places we want to go to, we don’t know the exact location.  A self-drive car does not know the exact location of the road so it looks at the print strips on the pavment and keep in the center of the lane.

On Nov 15, 2023, at 2:49 PM, Scott Monaghan <scott.m...@gmail.com> wrote:

[Marco Walther]

On 11/15/23 18:04, Chris Albertson wrote:
> This is why they stopped encrypting the GPS signal.   When GPS was new
> the US military used to encrypt the last few bits of the GPS data and
> without the key you’d only have an approximate location.   But then some
> smart guy said that if the GPS receiver were placed on a well-surveyed
> location it could know what those bits should be and could automatically
> tell anyone within radio range the correct value of the bits.  This made
> encryption pointless
>
> RTK is taking this same approach to propagation through the air.   It is
> random and makes GPS less accurate.  But a GPS at a surveyed point can
> measure the error and tell others.
>
> But really there are not a lot of uses for ultra-accurate GPS.  Most of
> the places we want to go to, we don’t know the exact location.  A
> self-drive car does not know the exact location of the road so it looks
> at the print strips on the pavment and keep in the center of the lane.

Agriculture, construction or geology are a few uses;-) I only saw the
first part of the article (don't have the needed account), but he is
using the well-known UBlox ZED-F9* dual band receivers and yes, they can
get down to cm level. My base station on the roof was 'post-processed'
to within 2mm. So, my lawn mower can normally navigate to within 20mm;-)

Iff you happen to come by a newly started 'commercial construction
side', the first couple of things, that go up are 'the office container'
and somewhere nearby (potentially on it's roof), the GNSS base
station;-) (A round plastic antenna, about 15cm diameter for the GNSS
receiver and another 'whip antenna' for the local broadcast. Iff you
then come back while bulldozers are working, you can find usually three
similar round antennas on them. One on the cabin (for the general
position and two on sticks above both sides of the shield. That way,
they can calculate the position of the shield in space to few millimeters.

With the UBlox boards, you can get that level for about $700-800 with
your own base station, half if you can use somebody else's;-) Five years
ago, the price was still around $5k for the same precision. You can
spend another $900 or so for a triple band setup which can give you few
millimeters in good conditions.

-- Marco

>
>> On Nov 15, 2023, at 2:49 PM, Scott Monaghan <scott.m...@gmail.com>
>> wrote:
>>
>> Hi all,
>>
>> My Twitter buddy Jason just posted this awesome article about a bot
>> that uses RTK GPS to achieve 14mm GPS accuracy with ROS2.
>>
>> Check out:

>> https://medium.com/exploring-ros-robotics/how-to-use-rtk-gps-on-a-ros-robot-a51e9aa2f2ab <https://medium.com/exploring-ros-robotics/how-to-use-rtk-gps-on-a-ros-robot-a51e9aa2f2ab>

>
> --
> You received this message because you are subscribed to the Google
> Groups "HomeBrew Robotics Club" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to hbrobotics+...@googlegroups.com

> <mailto:hbrobotics+...@googlegroups.com>.

> To view this discussion on the web visit

> https://groups.google.com/d/msgid/hbrobotics/67EEAE02-F832-4DA2-93AC-1A4F0FB9BB1E%40gmail.com <https://groups.google.com/d/msgid/hbrobotics/67EEAE02-F832-4DA2-93AC-1A4F0FB9BB1E%40gmail.com?utm_medium=email&utm_source=footer>.

[Marcos castro]

Hi there, I was looking for a way to apply location inside home to use with my robot, or even to develop any for blind people. Do you think this RTK GPS is a solution ? is it too expensive for a robot project? 

Marcos

--

You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.

To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/CAMzUCWOpsHiWvMGukrcF7VO4MMsMAdCN6ipfuneHkExyWP0eAA%40mail.gmail.com.

[Sergei Grichine]

@Marcos - RTK GPS is very dependent on a clear sky view, it doesn't work indoors.

@Chris - RTK GPS is different from "differential GPS", and relies on computing phase differences of the carrier signal. That's why its precision is comparable to the wavelength (L1, at 1575.42 MHz - 19cm, and L2, at 1227.6 MHz - 24cm). 

One of the advantages of RTK technology is that two RTK GPS receivers can compute true heading with great accuracy,

More info here (and, of course, in related Application Notes):

https://github.com/PX4/PX4-GPSDrivers/pull/131

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/CANw-TqAjO3U9vteKxFarMH5WqUQww%3DrFW_UqFQBiskRC0kAYxQ%40mail.gmail.com.

--

Best Regards,

-- Sergei

[camp .]

    I think accuracy is overrated when using GPS for navigation. Even if the error is greater than 10', it puts you in a room or a rough location and perhaps orientation within the environment. At this point, other sensors (camera, etc) can refine the localization. At the least, it gives a starting point for the kidnapped robot problem.

Thanks,

Camp

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/CA%2BKVXVOcPBH_vMmYfZyHPo-FNGL6%2BHnywo0yU0d92Y1LPqda9Q%40mail.gmail.com.

[Ralph Hipps]

technically, it's typed 14 mm, 24 cm, etc., not 14mm, 24cm.

just an fyi.     =)

[Chris Albertson]

On Nov 16, 2023, at 7:57 AM, camp . <ca...@camppeavy.com> wrote:

    I think accuracy is overrated when using GPS for navigation. Even if the error is greater than 10', it puts you in a room or a rough location and perhaps orientation within the environment. At this point, other sensors (camera, etc) can refine the localization. At the least, it gives a starting point for the kidnapped robot problem.

Very accurate GPS works best if you know the location you need to go to with similar accuracy.   This is generally not the case unless the robot always stays in one VERY well-mapped area.  Also, GPS in general does not work well indoors.   It can work in some wood-frame houses but not in places with substantial structure that blocks a clear view of the sky.

I think GPS is best used as one component of navigation.  Think of a car.  You do not want it to give up and stop driving if it goes into a tunnel of a parking structure that blocks GPS.  But GPS works well, when it works and you’d not want to give that up, so use it when you can.

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/233171234.4565713.1700150275546%40mail.yahoo.com.

[Sergei Grichine]

"In real life" robots don't use GPS alone, its data is combined with other location-related sensors (Accelerometer, Gyro, Magnetometer, Optical Flow, Pitot tubes, Odometry, Camera SLAM etc.). Here is a (stolen) simplified diagram:

To combine all data an EKF (Extended KalmanFilter) is commonly used. It computes state data (position, orientation, velocities). The less noise sensors can deliver, the better is output.

A drone can have several EKF instances with different settings, all competing for precision data that ultimately goes downstream to be used for Position Control.

It gets complicated real fast if the price of an error is high and precision requirements are also high. So, the 2 cm RTK precision and <1 degree heading accuracy is priceless.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/D1FA5944-0BFE-4652-AD47-DC25CD2CE785%40gmail.com.

[Danny Zemanek]

On Nov 16, 2023, at 11:24 AM, Sergei Grichine <vital...@gmail.com> wrote:

"In real life" robots don't use GPS alone, its data is combined with other location-related sensors (Accelerometer, Gyro, Magnetometer, Optical Flow, Pitot tubes, Odometry, Camera SLAM etc.). Here is a (stolen) simplified diagram:

To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/CA%2BKVXVOG69%3DagAhDj9uPFhT2djgSRN%2BtXq%2BSOFX5Y16Tv%3DJg3Q%40mail.gmail.com.

[Carl Sutter]

Another vendor of mostly ublox hardware + a number or articles: https://www.ardusimple.com/

Also, ubox sells "eval" boards for a number of their systems.