Hello, the organizers and others have mentioned that (combinatorial) Kalman filters are used in production, when trying to reconstruct tracks from hits.
\n\nMy short Google research indicated that this method is not suitable for our competition, because it runs in O(n^2) to O(n^3), where n is the number of hits (in our competition: 100.000 hits per event).
\n\nSo, Kalman would take too long. Instead, we would rather need an algorithm with around O(n * log(n)^2), at most (don't nail me down on the power of the logarithm though).
\n\nIs this correct?\nDoes anyone use the Kalman filter or an adjusted version?
", "messages": [ { "id": "336922", "postDate": "06/01/2018 14:41:55", "content": "Hello, the organizers and others have mentioned that (combinatorial) Kalman filters are used in production, when trying to reconstruct tracks from hits.
\n\nMy short Google research indicated that this method is not suitable for our competition, because it runs in O(n^2) to O(n^3), where n is the number of hits (in our competition: 100.000 hits per event).
\n\nSo, Kalman would take too long. Instead, we would rather need an algorithm with around O(n * log(n)^2), at most (don't nail me down on the power of the logarithm though).
\n\nIs this correct?\nDoes anyone use the Kalman filter or an adjusted version?
", "rawMarkdown": "Hello, the organizers and others have mentioned that (combinatorial) Kalman filters are used in production, when trying to reconstruct tracks from hits.\n\nMy short Google research indicated that this method is not suitable for our competition, because it runs in O(n^2) to O(n^3), where n is the number of hits (in our competition: 100.000 hits per event). \n\nSo, Kalman would take too long. Instead, we would rather need an algorithm with around O(n * log(n)^2), at most (don't nail me down on the power of the logarithm though).\n\nIs this correct?\nDoes anyone use the Kalman filter or an adjusted version?", "votes": null }, { "id": "338242", "postDate": "06/04/2018 17:23:11", "content": "Do you have a workable Kalman filter for this task?
", "rawMarkdown": "Do you have a workable Kalman filter for this task?", "votes": null }, { "id": "338264", "postDate": "06/04/2018 17:55:30", "content": "No - but since it is used by the pros in production, it makes super sense to have a look into this... unless it is computationally unfeasible, which is why I posted my question
", "rawMarkdown": "No - but since it is used by the pros in production, it makes super sense to have a look into this... unless it is computationally unfeasible, which is why I posted my question", "votes": null }, { "id": "339826", "postDate": "06/07/2018 18:24:59", "content": "I am looking into a simple model with kalman filter, it seems we only have limited information, with coordinates of (x, y, z) and q, any way we can create the system model of A, B and H?
", "rawMarkdown": "I am looking into a simple model with kalman filter, it seems we only have limited information, with coordinates of (x, y, z) and q, any way we can create the system model of A, B and H?", "votes": null }, { "id": "340319", "postDate": "06/08/2018 22:11:01", "content": "check this:
\n\nhttps://indico.nbi.ku.dk/event/758/attachments/1684/2344/trackingnote.pdf
", "rawMarkdown": "check this:\n\nhttps://indico.nbi.ku.dk/event/758/attachments/1684/2344/trackingnote.pdf", "votes": null } ], "comments": [ { "id": 338242, "author_name": "cosmojo", "author_url": "", "post_date": "06/04/2018 17:23:11", "content": "Do you have a workable Kalman filter for this task?
", "votes": null, "replies": [ { "id": 338264, "author_name": "trian2018", "author_url": "", "post_date": "06/04/2018 17:55:30", "content": "No - but since it is used by the pros in production, it makes super sense to have a look into this... unless it is computationally unfeasible, which is why I posted my question
", "votes": null, "replies": [] } ] }, { "id": 339826, "author_name": "cosmojo", "author_url": "", "post_date": "06/07/2018 18:24:59", "content": "I am looking into a simple model with kalman filter, it seems we only have limited information, with coordinates of (x, y, z) and q, any way we can create the system model of A, B and H?
", "votes": null, "replies": [ { "id": 340319, "author_name": "hengck23", "author_url": "", "post_date": "06/08/2018 22:11:01", "content": "check this:
\n\nhttps://indico.nbi.ku.dk/event/758/attachments/1684/2344/trackingnote.pdf
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "336922": "Hello, the organizers and others have mentioned that (combinatorial) Kalman filters are used in production, when trying to reconstruct tracks from hits.\n\nMy short Google research indicated that this method is not suitable for our competition, because it runs in O(n^2) to O(n^3), where n is the number of hits (in our competition: 100.000 hits per event). \n\nSo, Kalman would take too long. Instead, we would rather need an algorithm with around O(n * log(n)^2), at most (don't nail me down on the power of the logarithm though).\n\nIs this correct?\nDoes anyone use the Kalman filter or an adjusted version?", "338242": "Do you have a workable Kalman filter for this task?", "338264": "No - but since it is used by the pros in production, it makes super sense to have a look into this... unless it is computationally unfeasible, which is why I posted my question", "339826": "I am looking into a simple model with kalman filter, it seems we only have limited information, with coordinates of (x, y, z) and q, any way we can create the system model of A, B and H?", "340319": "check this:\n\nhttps://indico.nbi.ku.dk/event/758/attachments/1684/2344/trackingnote.pdf" }, "source": "meta" }