Repository navigation
Bosch-A brakes: smoother stop-and-go, and no hard curve brake behind a lead that isn't slowing - #21
Closed
trung791997 wants to merge 3 commits into
Closed
trung791997 wants to merge 3 commits into
trung791997 wants to merge 3 commits into
Conversation
…ake behind a lead that isn't slowing Stop-and-go on the Civic: the brake no longer switches on and off 15-25 times a minute at 2-4 m/s, and the gas no longer jumps to the 60 cap the moment the brake lets go. The brake request now starts below -0.20 (was -0.12), stays on at least 1.5 s unless real acceleration is asked for, and gas after a release starts at 35 and ramps to the lookup over 1 s. Runs on Bosch-A radar harness cars (Civic Bosch, CR-V 5G); the onset compensation now follows the same gate. Replay of route 0000001f (open loop): toggles 3.69 -> 3.17/min, holds under 1.5 s 27.6 -> 16.7 %, gas on the release frame 60 -> 35. On Peter's Civic since 2026-10-24 (limited road evidence); CR-V static only. Curve brake: on route 30 an S-curve gave a -2.87 brake behind a lead that was not slowing; the radar track read +4 then -3 m/s^2 within a second while the camera saw a steady car. For 4 s after a radar lead's accel flips from above +2 to below -2 within 1.5 s, its speed and accel move 60 % toward the camera lead, only in the less-severe direction and only when the camera sees a lead. Distance is never changed and no point is dropped. No bound under 30 m or TTC 4 s. Bosch-A only, default on. Log evidence across 18 routes; on Peter's Civic since 2026-10-26 (limited road evidence). Planner: the Human Acceleration cap is applied before the pre-AP cap (braking targets pass through both unchanged). Static: Honda opendbc tests, starpilot Honda tests and the planner/radar tests, 1536 passed. ruff clean. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nc dropped CI process replay stopped before comparing anything, because the reviewed-reference manifest names two files that were not in the tree: regen218A4DCFAA1_2025-04-08--22-57-51--0_plannerd.zst regen7366F13F6A1_2025-04-08--23-07-07--0_card.zst *.zst is gitignored, so the sync at b47e51c left them out. Both are restored byte for byte from 3420f74; their sha256 and sizes match manifest.json. No manifest or code change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Owner
|
Re-running CI against night-star7 5dc8f5f (restored files + FLM test fixes); reopening. |
Owner
|
Re-running CI against night-star7 386bb02 (static-analysis and connect-test fixes on the base); reopening. |
…is, FLM tests, restored upstream files) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
JamesL787
pushed a commit
that referenced
this pull request
Oct 31, 2026
process replay: reviewed references for the four intentional Honda-PTM differences
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.
Stop-and-go on the Civic: the brake no longer switches on and off 15-25 times a minute at 2-4 m/s,
and the gas no longer jumps to the 60 cap the moment the brake lets go. The brake request now
starts below -0.20 (was -0.12), stays on at least 1.5 s unless real acceleration is asked for,
and gas after a release starts at 35 and ramps to the lookup over 1 s. Runs on Bosch-A radar
harness cars (Civic Bosch, CR-V 5G); the onset compensation now follows the same gate.
Replay of route 0000001f (open loop): toggles 3.69 -> 3.17/min, holds under 1.5 s 27.6 -> 16.7 %,
gas on the release frame 60 -> 35. On Peter's Civic since 2026-10-24 (limited road evidence);
CR-V static only.
Curve brake: on route 30 an S-curve gave a -2.87 brake behind a lead that was not slowing; the
radar track read +4 then -3 m/s^2 within a second while the camera saw a steady car. For 4 s
after a radar lead's accel flips from above +2 to below -2 within 1.5 s, its speed and accel move
60 % toward the camera lead, only in the less-severe direction and only when the camera sees a
lead. Distance is never changed and no point is dropped. No bound under 30 m or TTC 4 s.
Bosch-A only, default on. Log evidence across 18 routes; on Peter's Civic since 2026-10-26
(limited road evidence).
Planner: the Human Acceleration cap is applied before the pre-AP cap (braking targets pass
through both unchanged).
Static: Honda opendbc tests, starpilot Honda tests and the planner/radar tests, 1536 passed. ruff clean.
Not included: the radard aLeadK clamp / aLeadTau floor (already on night-star7), and the far brake-tap fix (radar_lead_bound), which follows after its first drive.
🤖 Generated with Claude Code