{ "id": 251463, "title": "Is there anyone using the \"optional supplemental datasets\"?", "url": "/competitions/google-smartphone-decimeter-challenge/discussion/251463", "author_name": "", "post_date": "2021-07-07T12:25:46.722698200Z", "votes": 7, "comment_count": 6, "views": 0, "content": "

For the past week,
\nI've been investigated using \"optional supplemental datasets\" to improve baseline.
\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579

\n

I considered using the osr data from this repository. I aimed to correct the pseudorange using observation data from a base station whose position was known.

\n

I have succeeded in decoding the json data and merging the observation data with the ephemeris data. However, I'm having difficulties in calculating the distance to the base station by calculating the position of the satellite from the ephemeris data, but this number is not reasonable…

\n

Has anyone considered (and succeeded in) a similar approach?

", "messages": [ { "id": "1379509", "postDate": "07/07/2021 12:25:46", "content": "

For the past week,
\nI've been investigated using \"optional supplemental datasets\" to improve baseline.
\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579

\n

I considered using the osr data from this repository. I aimed to correct the pseudorange using observation data from a base station whose position was known.

\n

I have succeeded in decoding the json data and merging the observation data with the ephemeris data. However, I'm having difficulties in calculating the distance to the base station by calculating the position of the satellite from the ephemeris data, but this number is not reasonable…

\n

Has anyone considered (and succeeded in) a similar approach?

", "rawMarkdown": "For the past week, \nI've been investigated using \"optional supplemental datasets\" to improve baseline.\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579\n\nI considered using the osr data from [this repository](https://github.com/swift-nav/google-sdc-corrections). I aimed to correct the pseudorange using observation data from a base station whose position was known. \n\nI have succeeded in decoding the json data and merging the observation data with the ephemeris data. However, I'm having difficulties in calculating the distance to the base station by calculating the position of the satellite from the ephemeris data, but this number is not reasonable... \n\nHas anyone considered (and succeeded in) a similar approach?", "votes": null }, { "id": "1380167", "postDate": "07/07/2021 21:50:32", "content": "

Yes, I've succeeded in calculating the satellite positions from the ephemeris files - where are you getting stuck? I can try to help or point to resources.

\n

Edit: I suppose I can give a starting point for others looking at this: the SwiftNav documentation points to IS-GPS-200H (here: https://www.gps.gov/technical/icwg/IS-GPS-200H.pdf) for how to calculate satellite positions. Start around page 102, and match up the values from the swiftnav files to those values in the table, and then follow the calculations following that.

\n

It's long and a bit tricky, but all the info is there.

\n

I will also say: although the satellite positions that I calculated looked good (very close to the given locations), my accuracy actually went down when I used the corrected positions… so I'm actually not sure if I 100% got the correct calculations, or if I (more likely) am off in my timing by a few nanoseconds or something similar…

", "rawMarkdown": "Yes, I've succeeded in calculating the satellite positions from the ephemeris files - where are you getting stuck? I can try to help or point to resources.\n\nEdit: I suppose I can give a starting point for others looking at this: the SwiftNav documentation points to IS-GPS-200H (here: https://www.gps.gov/technical/icwg/IS-GPS-200H.pdf) for how to calculate satellite positions. Start around page 102, and match up the values from the swiftnav files to those values in the table, and then follow the calculations following that. \n\nIt's long and a bit tricky, but all the info is there.\n\nI will also say: although the satellite positions that I calculated looked good (very close to the given locations), my accuracy actually went _down_ when I used the corrected positions... so I'm actually not sure if I 100% got the correct calculations, or if I (more likely) am off in my timing by a few nanoseconds or something similar...", "votes": null }, { "id": "1380800", "postDate": "07/08/2021 10:15:34", "content": "

Thank you! You are very kind.
\nThe material you gave me is very easy to understand. I will try it out.

\n
\n

my accuracy actually went down when I used the corrected positions…

\n
\n

Thanks for sharing!
\nThis may mean that the error that can be corrected by DGNSS is not so important.

\n

BTW,
\nYour score has improved dramatically!
\nI often see people improving their scores significantly in this competition.
\nI'm hoping to find some solutions too.

", "rawMarkdown": "Thank you! You are very kind. \nThe material you gave me is very easy to understand. I will try it out.\n\n> my accuracy actually went down when I used the corrected positions… \n\nThanks for sharing! \nThis may mean that the error that can be corrected by DGNSS is not so important.\n\nBTW,\nYour score has improved dramatically! \nI often see people improving their scores significantly in this competition. \nI'm hoping to find some solutions too.", "votes": null }, { "id": "1380916", "postDate": "07/08/2021 12:38:39", "content": "

No problem!

\n

And yes, since I finally got my gps code all working, I've been able to improve a lot :)

", "rawMarkdown": "No problem!\n\nAnd yes, since I finally got my gps code all working, I've been able to improve a lot :)", "votes": null }, { "id": "1383461", "postDate": "07/10/2021 21:16:00", "content": "

Hello @chris62 , I am new to GPS topics. I am also not following what you try to do..are you trying to improve baseline lat and long ?How do you use supplement data to improve things?

", "rawMarkdown": "Hello @chris62 , I am new to GPS topics. I am also not following what you try to do..are you trying to improve baseline lat and long ?How do you use supplement data to improve things?", "votes": null }, { "id": "1383520", "postDate": "07/11/2021 00:33:40", "content": "

There are (at least) two ways you might use supplemental datasets (at least for the ones listed here: https://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579)

\n
    \n

  1. Improving satellite positions: that is what this thread is about. The satellite positions given in the _derived file are very close to correct, but there are additional corrections given by certain sources (including a couple listed in the link above). Using that data, you can predict slightly better satellite positions, which should (in theory) improve the baseline accuracy.

    2. \n

  2. Comparing with GPS reference stations: there are several gps reference stations that provide data that you can compare with the \"rover\" data (phone data) to do differential, or relative gps. That's not what this thread was originally about, but it's the other major use of supplemental data sets.

    3. \n
\n

That said: It took me quite a while to understand the supplemental datasets, but there is a lot of information online if you look for it. But it is definitely not straightforward to apply the supplemental data in an easy way - it takes a bit of work to get it all working correctly.

", "rawMarkdown": "There are (at least) two ways you might use supplemental datasets (at least for the ones listed here: https://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579)\n\n1. Improving satellite positions: that is what this thread is about. The satellite positions given in the _derived file are very close to correct, but there are additional corrections given by certain sources (including a couple listed in the link above). Using that data, you can predict slightly better satellite positions, which should (in theory) improve the baseline accuracy.\n\n2. Comparing with GPS reference stations: there are several gps reference stations that provide data that you can compare with the \"rover\" data (phone data) to do differential, or relative gps. That's not what this thread was originally about, but it's the other major use of supplemental data sets.\n\nThat said: It took me quite a while to understand the supplemental datasets, but there is a lot of information online if you look for it. But it is definitely not straightforward to apply the supplemental data in an easy way - it takes a bit of work to get it all working correctly.", "votes": null }, { "id": "1384270", "postDate": "07/11/2021 16:06:30", "content": "

Thanks for detailed explanation @chris62

", "rawMarkdown": "Thanks for detailed explanation @chris62", "votes": null } ], "comments": [ { "id": 1380167, "author_name": "chris62", "author_url": "", "post_date": "07/07/2021 21:50:32", "content": "

Yes, I've succeeded in calculating the satellite positions from the ephemeris files - where are you getting stuck? I can try to help or point to resources.

\n

Edit: I suppose I can give a starting point for others looking at this: the SwiftNav documentation points to IS-GPS-200H (here: https://www.gps.gov/technical/icwg/IS-GPS-200H.pdf) for how to calculate satellite positions. Start around page 102, and match up the values from the swiftnav files to those values in the table, and then follow the calculations following that.

\n

It's long and a bit tricky, but all the info is there.

\n

I will also say: although the satellite positions that I calculated looked good (very close to the given locations), my accuracy actually went down when I used the corrected positions… so I'm actually not sure if I 100% got the correct calculations, or if I (more likely) am off in my timing by a few nanoseconds or something similar…

", "votes": null, "replies": [ { "id": 1380800, "author_name": "t88take", "author_url": "", "post_date": "07/08/2021 10:15:34", "content": "

Thank you! You are very kind.
\nThe material you gave me is very easy to understand. I will try it out.

\n
\n

my accuracy actually went down when I used the corrected positions…

\n
\n

Thanks for sharing!
\nThis may mean that the error that can be corrected by DGNSS is not so important.

\n

BTW,
\nYour score has improved dramatically!
\nI often see people improving their scores significantly in this competition.
\nI'm hoping to find some solutions too.

", "votes": null, "replies": [] }, { "id": 1380916, "author_name": "chris62", "author_url": "", "post_date": "07/08/2021 12:38:39", "content": "

No problem!

\n

And yes, since I finally got my gps code all working, I've been able to improve a lot :)

", "votes": null, "replies": [] } ] }, { "id": 1383461, "author_name": "prashantkramadhari", "author_url": "", "post_date": "07/10/2021 21:16:00", "content": "

Hello @chris62 , I am new to GPS topics. I am also not following what you try to do..are you trying to improve baseline lat and long ?How do you use supplement data to improve things?

", "votes": null, "replies": [ { "id": 1383520, "author_name": "chris62", "author_url": "", "post_date": "07/11/2021 00:33:40", "content": "

There are (at least) two ways you might use supplemental datasets (at least for the ones listed here: https://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579)

\n
    \n

  1. Improving satellite positions: that is what this thread is about. The satellite positions given in the _derived file are very close to correct, but there are additional corrections given by certain sources (including a couple listed in the link above). Using that data, you can predict slightly better satellite positions, which should (in theory) improve the baseline accuracy.

    2. \n

  2. Comparing with GPS reference stations: there are several gps reference stations that provide data that you can compare with the \"rover\" data (phone data) to do differential, or relative gps. That's not what this thread was originally about, but it's the other major use of supplemental data sets.

    3. \n
\n

That said: It took me quite a while to understand the supplemental datasets, but there is a lot of information online if you look for it. But it is definitely not straightforward to apply the supplemental data in an easy way - it takes a bit of work to get it all working correctly.

", "votes": null, "replies": [] }, { "id": 1384270, "author_name": "prashantkramadhari", "author_url": "", "post_date": "07/11/2021 16:06:30", "content": "

Thanks for detailed explanation @chris62

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1379509": "For the past week, \nI've been investigated using \"optional supplemental datasets\" to improve baseline.\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579\n\nI considered using the osr data from [this repository](https://github.com/swift-nav/google-sdc-corrections). I aimed to correct the pseudorange using observation data from a base station whose position was known. \n\nI have succeeded in decoding the json data and merging the observation data with the ephemeris data. However, I'm having difficulties in calculating the distance to the base station by calculating the position of the satellite from the ephemeris data, but this number is not reasonable... \n\nHas anyone considered (and succeeded in) a similar approach?", "1380167": "Yes, I've succeeded in calculating the satellite positions from the ephemeris files - where are you getting stuck? I can try to help or point to resources.\n\nEdit: I suppose I can give a starting point for others looking at this: the SwiftNav documentation points to IS-GPS-200H (here: https://www.gps.gov/technical/icwg/IS-GPS-200H.pdf) for how to calculate satellite positions. Start around page 102, and match up the values from the swiftnav files to those values in the table, and then follow the calculations following that. \n\nIt's long and a bit tricky, but all the info is there.\n\nI will also say: although the satellite positions that I calculated looked good (very close to the given locations), my accuracy actually went _down_ when I used the corrected positions... so I'm actually not sure if I 100% got the correct calculations, or if I (more likely) am off in my timing by a few nanoseconds or something similar...", "1380800": "Thank you! You are very kind. \nThe material you gave me is very easy to understand. I will try it out.\n\n> my accuracy actually went down when I used the corrected positions… \n\nThanks for sharing! \nThis may mean that the error that can be corrected by DGNSS is not so important.\n\nBTW,\nYour score has improved dramatically! \nI often see people improving their scores significantly in this competition. \nI'm hoping to find some solutions too.", "1380916": "No problem!\n\nAnd yes, since I finally got my gps code all working, I've been able to improve a lot :)", "1383461": "Hello @chris62 , I am new to GPS topics. I am also not following what you try to do..are you trying to improve baseline lat and long ?How do you use supplement data to improve things?", "1383520": "There are (at least) two ways you might use supplemental datasets (at least for the ones listed here: https://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/238579)\n\n1. Improving satellite positions: that is what this thread is about. The satellite positions given in the _derived file are very close to correct, but there are additional corrections given by certain sources (including a couple listed in the link above). Using that data, you can predict slightly better satellite positions, which should (in theory) improve the baseline accuracy.\n\n2. Comparing with GPS reference stations: there are several gps reference stations that provide data that you can compare with the \"rover\" data (phone data) to do differential, or relative gps. That's not what this thread was originally about, but it's the other major use of supplemental data sets.\n\nThat said: It took me quite a while to understand the supplemental datasets, but there is a lot of information online if you look for it. But it is definitely not straightforward to apply the supplemental data in an easy way - it takes a bit of work to get it all working correctly.", "1384270": "Thanks for detailed explanation @chris62" }, "source": "meta" }