research: cave planner merge evidence (L150, measurement only)
Merges cleanly. UE start: S1 -9%, S3 +11%, S0/S2 ~+1%. 30 m/s: near coverage 44.2 -> 38.5%, far holes +0.06 pp, p99 +2%. L80 near-ready slower in 6/8 cells (up to +7.9%), whole view faster in 7/8. Sweeps 15/16 each (seed 2). Source/: VoxelForgeCoreAdapter.cpp. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016dhxM43dnWvUu2B5Ph1iG1
This commit is contained in:
1 parent
b8382eed84
commit
9ac60d5d18
15 files changed
+871
No files matched your search
@@ -0,0 +1,14 @@
|
||||
# L150 matched owner-world captures
|
||||
|
||||
Seed 0, with camera poses from the planner layout resolved against the owner's embedded Core plan and reused for both versions. The plan is imported without a seed override. The two Tunnel Network views sit at the planned route and departure joints; the Vertical Shafts views look down from the access landing and up from a ledge.
|
||||
|
||||
- Tunnel Network / tunnel-network-route: [assembler-world](assembler-world-tunnel-network-route.jpg) — FAIL
|
||||
- Tunnel Network / tunnel-network-departure: [assembler-world](assembler-world-tunnel-network-departure.jpg) — PASS
|
||||
- Vertical Shafts / vertical-shafts-looking-down: [assembler-world](assembler-world-vertical-shafts-looking-down.jpg) — FAIL
|
||||
- Vertical Shafts / vertical-shafts-looking-up: [assembler-world](assembler-world-vertical-shafts-looking-up.jpg) — FAIL
|
||||
- Tunnel Network / tunnel-network-route: [planner-head](planner-head-tunnel-network-route.jpg) — PASS
|
||||
- Tunnel Network / tunnel-network-departure: [planner-head](planner-head-tunnel-network-departure.jpg) — PASS
|
||||
- Vertical Shafts / vertical-shafts-looking-down: [planner-head](planner-head-vertical-shafts-looking-down.jpg) — PASS
|
||||
- Vertical Shafts / vertical-shafts-looking-up: [planner-head](planner-head-vertical-shafts-looking-up.jpg) — PASS
|
||||
|
||||
All eight exports and renders completed. The capture-quality gate passed for all four planner-head images and one assembler/world image. Three assembler/world images have insufficient image variance because no mesh surface is visible from those same viewpoints; those gate failures are recorded as the baseline result.
|
||||
Reference in new issue
Block a user