Parallel assumption and complex topologies about lstr HOT 1 CLOSED

Hi @lz707 ,
Thanks for your interest and sorry for getting back to you so late!

Yes, your understanding is correct.
The current scenario considered in our method is one in which the vehicle is driving steadily on the road and the road is clearly marked with lane markings.

Merges and splits are a great follow-up. In this scenario the lane patterns get complicated and the global parallel assumptions obviously cause problems.
For example, there are 4 lanes near the freeway ramp, and the left 2 lanes are straight while the 2 on the right are curved. We think

1. a more flexible local parallel assumption that can automatically cluster lanes of the same shape.
2. a more diverse dataset containing the aforementioned lane patterns is also critical.

Some brief insights and I welcome in-depth discussion!

from lstr.