Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Speedy Bus Trip - Data Anamoly #1099

Open
Abby-Wheelis opened this issue Dec 9, 2024 · 10 comments
Open

Speedy Bus Trip - Data Anamoly #1099

Abby-Wheelis opened this issue Dec 9, 2024 · 10 comments

Comments

@Abby-Wheelis
Copy link
Member

In ongoing analysis, we found a bus trip with extremely high emissions because it had a high distance, over 1,600 miles in about 2 hours. The start and end locs are only about 43 miles or 55 min by car apart.

This outlier appears to be skewing the total bus emissions for the program.

We are investigating to see what happened with this trip and why it was tracked as if it was taken at the speed of light.

The trip trajectory shows the trip between the start and end as a straight line, with a long straight line trip halfway across the country in between. The only things we can think of are opcode sharing, magic, or some extreme tracking error. It might be several trips that did not get segmented properly?

Looking at the time range of the trajectory points, it was around 2 hours and 15 minutes. So still fast, and maybe tracked wrong, but not speed of light fast.

@JGreenlee and @jpfleischer for visibility

@shankari
Copy link
Contributor

shankari commented Dec 9, 2024

That sounds like a tracking error that our smoothing algorithms missed. Our smoothing algorithms are definitely not perfect!

Can we look at the points right around the jump? In general, with shared opcodes, we will get points A, B, C in loc1 and then points A', B', C' in loc2 or some interleaving. But if there is a jump from A, B, C in loc1 to a small number of points (< 10-15) in loc2 and then a jump back to loc1 and the user doesn't go back to loc2 that day, it is almost certainly a tracking error (try to think of a scenario where that could happen IRL; I can't 😄 )

@Abby-Wheelis
Copy link
Member Author

All of the tracking was straight line, with intermediary points. The map essentially looks like a set of dotted lines - one between the start and end point of the trip, one from that area to about 1,000 miles away, and then another short dotted line in the area at the end of the long trajectory. It could be a ride to the airport, a flight, and then a ride home? But with lots of errors in the smoothing that led to the timing issue?

Looking into the sequence of the dots now...

@Abby-Wheelis
Copy link
Member Author

Looking at the sequence, the short trip from the recorded start to end happened first, then the short trip, 1,000 miles away, then the trip from 1,000 miles away to approx the same place as the end of the trip record. Maybe 1 user took the recorded trip, and the 2nd user drove to the airport 1,000 miles away then flew to the same end location. If it happened around the same time our algorithms just got very thrown off.

If all location points were from the same person, they would have had to have taken a short trip, magically moved 1,000 miles in an instant, taken another short trip, then flown back home. Even though all the points are in sequence (no hopping back and forth) it must have been at least two people responsible for recording the points.

@shankari
Copy link
Contributor

shankari commented Dec 9, 2024

it must have been at least two people responsible for recording the points.

Or it is just GPS noise. I don't know if you have seen the prior examples of big jumps
e.g. #843

although that was easier because the jump was to the middle of the ocean

@JGreenlee
Copy link

JGreenlee commented Dec 9, 2024

If it's just GPS noise, why are there intermediate points along the path from Minneapolis back to Colorado Springs?

@Abby-Wheelis
Copy link
Member Author

Can we post the trajectory here since it is not associated with a user? There is a graph of the points color-scaled according to time which is where I observed the sequence

@shankari
Copy link
Contributor

If it's just GPS noise, why are there intermediate points along the path from Minneapolis back to Colorado Springs?

Because the noise characteristics are not always predictable.
See #843
there was a clearly incorrect jump to just off the coast of Africa.
and we actually received multiple points in the incorrect location.

@shankari
Copy link
Contributor

Can we post the trajectory here since it is not associated with a user? There is a graph of the points color-scaled according to time which is where I observed the sequence

I think it is fine to post as long as we are focused on the noisy part. Or create an internal issue and post.

you might also want to look at the speed of the jump. One of the clues for 843 (other than the fact that we ended up in the middle of the ocean) was that the speed was ~ Mach 2.

IIRC, with the retirement of the concorde, there are currently no commercial aircraft that fly at supersonic speeds.

@Abby-Wheelis
Copy link
Member Author

I think it is fine to post as long as we are focused on the noisy part. Or create an internal issue and post.

There isn't really a noisy part to zoom in on, there are points all along the way of all three "segments", I added num 58 internally to show the plot of points.

@JGreenlee
Copy link

If it's just GPS noise, why are there intermediate points along the path from Minneapolis back to Colorado Springs?

Because the noise characteristics are not always predictable. See #843 there was a clearly incorrect jump to just off the coast of Africa. and we actually received multiple points in the incorrect location.

I'm not convinced it is the same phenomenon. #843 just looked like an isolated cluster of points near Null Island.

In this case, there was actually a steady stream of points along the entire return path (from Minnesota back to Pueblo).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants