Debug clustering issues
Description
In !1174 (comment 672788198) we see that clustering behaves in a suboptimal way. In this issue we should look into why that is the case and if this can be improved.
Currently found issues:
- As discussed in #1322 (comment 675761680) if the car is not exactly parallel to the ground, our ground detection algorithms fails
- As discussed in #1322 (comment 675794690) in certain areas of the environment (huge walls surrounding the ego vehicle) we get too many points to cluster, which eventually takes too long, thus there is lag.
Definition of Done
-
It is clear why clustering behaves badly in examples from the link above -
A fix or a better set of parameters are found to be used for the tracking demo
Edited by Igor Bogoslavskyi