{
  "id": 177944,
  "title": "Simple constant velocity extrapolation baseline",
  "url": "/competitions/lyft-motion-prediction-autonomous-vehicles/discussion/177944",
  "author_name": "ryches",
  "post_date": "2020-08-27T23:49:00.583000",
  "votes": 16,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I made a simple kernel showing how to make the simplest possible extrapolations just assuming constant velocity from the vehicle. The logic here is that a car will likely continue along the road with the same speed and heading as the last known points. </p>\n<p>You can calculate the delta of x and y of the last two known points and then continue adding that difference to the last known point to apply the same vector into the future 50 timesteps we are predicting.</p>\n<p>Obviously this isn't very sophisticated and doesn't take into account bends in the road or intersections where decisions are made and turns are possible, but it at least places a score to what reasonably calibrated predictions should yield. </p>\n<p><a href=\"https://www.kaggle.com/ryches/lyft-constant-velocity-extrapolation-baseline/notebook\" target=\"_blank\">https://www.kaggle.com/ryches/lyft-constant-velocity-extrapolation-baseline/notebook</a></p>",
  "messages": [
    {
      "id": 988267,
      "postDate": "2020-08-27T23:49:00.583Z",
      "content": "<p>I made a simple kernel showing how to make the simplest possible extrapolations just assuming constant velocity from the vehicle. The logic here is that a car will likely continue along the road with the same speed and heading as the last known points. </p>\n<p>You can calculate the delta of x and y of the last two known points and then continue adding that difference to the last known point to apply the same vector into the future 50 timesteps we are predicting.</p>\n<p>Obviously this isn't very sophisticated and doesn't take into account bends in the road or intersections where decisions are made and turns are possible, but it at least places a score to what reasonably calibrated predictions should yield. </p>\n<p><a href=\"https://www.kaggle.com/ryches/lyft-constant-velocity-extrapolation-baseline/notebook\" target=\"_blank\">https://www.kaggle.com/ryches/lyft-constant-velocity-extrapolation-baseline/notebook</a></p>",
      "rawMarkdown": "I made a simple kernel showing how to make the simplest possible extrapolations just assuming constant velocity from the vehicle. The logic here is that a car will likely continue along the road with the same speed and heading as the last known points. \n\nYou can calculate the delta of x and y of the last two known points and then continue adding that difference to the last known point to apply the same vector into the future 50 timesteps we are predicting.\n\nObviously this isn't very sophisticated and doesn't take into account bends in the road or intersections where decisions are made and turns are possible, but it at least places a score to what reasonably calibrated predictions should yield. \n\nhttps://www.kaggle.com/ryches/lyft-constant-velocity-extrapolation-baseline/notebook",
      "votes": 15
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "988267": "I made a simple kernel showing how to make the simplest possible extrapolations just assuming constant velocity from the vehicle. The logic here is that a car will likely continue along the road with the same speed and heading as the last known points. \n\nYou can calculate the delta of x and y of the last two known points and then continue adding that difference to the last known point to apply the same vector into the future 50 timesteps we are predicting.\n\nObviously this isn't very sophisticated and doesn't take into account bends in the road or intersections where decisions are made and turns are possible, but it at least places a score to what reasonably calibrated predictions should yield. \n\nhttps://www.kaggle.com/ryches/lyft-constant-velocity-extrapolation-baseline/notebook"
  }
}