Modify uncertainty reward so sensors measure all tracks together #691
Conversation
This can be important in case of sensor handling obscuration, or limited resolution. It also offers a small performance benefit.
Codecov Report
@@ Coverage Diff @@
## main #691 +/- ##
==========================================
- Coverage 94.62% 94.62% -0.01%
==========================================
Files 171 171
Lines 8576 8574 -2
Branches 1663 1662 -1
==========================================
- Hits 8115 8113 -2
Misses 339 339
Partials 122 122
Flags with carried forward coverage won't be shown. Click here to find out more.
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. |
| predicted_track.states = copy.copy(predicted_track.states) | ||
| predicted_track.metadatas = copy.copy(predicted_track.metadatas) |
There was a problem hiding this comment.
Why do you need to copy everything separately here and why are they copies of themselves? (Probably just me not understanding!)
There was a problem hiding this comment.
When you append states it'll modify the original list (and metadatas list), which you don't want. But if you deep copy the entire track to avoid this, it'll also copy all the states of the track which isn't required and super slow for any track with a lengthy history.
I should add a comment to make this clear in the code.
There was a problem hiding this comment.
In fact, having thought about this, I think we should implement a proper shallow copy method like a list/dictionary would have. I'll create a new PR for this once this is merged. It'll be same as here effectively, but part of the class, so a plain copy.copy(track) will ensure you get a shallow copy, and adding/removing states won't effect the original.
This can be important in case of sensor handling obscuration, or limited resolution. It also offers a small performance benefit.