-
Notifications
You must be signed in to change notification settings - Fork 1.3k
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
Inexplicable Bolus #2166
Comments
I looked at this a bit and I think this is probably a legitimate bug. The logs show that the user was in open loop all night. The insulin effects on the prediction eventually run out, at which point the RC terms keep the prediction extended out about an hour. So far, the the automaticDoseRecommendation is well behaved (less than 1U and consistent with ISF to bring prediction into range). Next, the RC prediction drops out and the prediction becomes only 15 minutes long with Momentum as the only contributor. At this time, the automaticDoseRecommendation jumps to 6U (40% of max bolus of 15U). The predictions and ISF don't support this dose at all. It's significant that the prediction is well-behaved, but the recommendation makes no sense.
The erroneous recommendations continue for almost 2 hours until the user goes back to closed loop and the 6U dose is actually delivered (and luckily interrupted by the user). After that, the recommendations go back to making sense. settings.json shows the open and closed loop timing as described above. However, the user does not recall closing the loop in the morning and doesn't remember going open loop for the night, but did recall having gone open loop for a while during the previous day (see FB discussion). I'm inclined to believe settings.json here, but mention this for completeness. I attempted to reproduce this result on my backup phone a few times by entering insulin in the past and running open loop until the insulin and RC effects dropped out. I achieved the shortened forecasts that match the issue report, but didn't trigger an erroneous bolus in this very limited testing. |
There is a somewhat similar report on Zulip here. In this account, user HEDRG got a large bolus after being away from his RileyLink all night. The logs aren't as useful because there aren't any predictions during the outage so we don't get to see if the automaticDoseRecommendation goes bad at exactly the time that the prediction becomes Momentum-only, but there are some strong similarities:
It's worth mentioning that this user had extensive code customizations including shortening insulin duration significantly. |
Describe the bug
Loop automatically bolused an amount that is not understood.
This may be related to Issue #2156
I hope @ps2 can look at these 2 issues.
Facebook Link:
Summary:
Comments:
Graphics from FB:
Details:
These are the 2 adjacent records from Doses.json that show the gap.
Build Details - Mac build
hardware:
uploaded files:
Export-20240505T133939Z.zip
Loop Report 2024-05-05 16_38_35+03_00.md
The text was updated successfully, but these errors were encountered: