Conversation
HansRobo
force-pushed
the
pr/scenario-sim-scene-trace
branch
2 times, most recently
from
August 21, 2026 06:40
b7b2fa3 to
961631f
Compare
A rollout writes what happened as geometry rather than as pixels: one gzipped JSONL frame per tick with the pre-step scene and the map-frame plan, plus one map asset per map shared by every case on it. It imports no matplotlib and invokes no encoder, so a run that draws nothing still reports what happened. Warm-up ticks carry null metrics: nothing scored them.
HansRobo
force-pushed
the
pr/scenario-sim-scene-trace
branch
from
August 27, 2026 06:59
961631f to
6e78c24
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
A
scenario_simrollout can report what happened as geometry instead of as pixels.Per tick it appends one frame to
scene_trace.jsonl.gz: the pre-step scene and the map-frameplan the renderer is handed, with clearance and collision attached once the ego's history is warm
(warm-up ticks carry nulls, because nothing scored them). Per map it writes one asset, shared by
every case on that map, into the run directory rather than the case directory.
It imports no matplotlib and invokes no encoder, so a run that draws nothing still reports what
happened.
RolloutConfig.write_scene_tracegates it.Why
PNG or mp4 output answers "what did this look like" but not "where was the ego, what was around it,
and what did it plan". The trace is small enough to keep for a whole suite and is what a browser
replay reads.
Note for review
write_scene_tracecurrently defaults toTrue, so every run starts writingscene_trace.jsonl.gzand a map asset. Happy to flip it to opt-in if that default is not wanted.Tests
tests/test_scene_trace.py(new): the frame the writer emits, the warm-up nulls, and the mapasset's placement and reuse.
scenario_generation/tests/: 257 passed / 15 skipped / 2 failed. The two failures aretest_map_cache.py, which is red ontier4-mainitself and is fixed by #48.