Add a distance-integrated steering-curvature correction to the restored v1
heading request. Hold the correction at zero error, allow unwind at limits,
and clear it for driver override or inactive PSCM control. Preserve C0,
C2=C3=0, final output limits, the 100 Hz sender and the existing opt-in toggle.
Validate build/hold/unwind, measurement cadence, anti-windup and the actual
controlsd-to-CAN path. The combined suite passes 511 tests and 9,146 subtests;
178 inherited or unsupported safety variants skip. Stress and frozen b8
replay pass 578,569 Float32/CAN round trips. These checks do not establish
physical tracking or stability. Record reproduction steps and source hashes.
Restore the exact tracked tree from 5fc16abc7, identified in the
9b and 9e strong-tracking routes, including opendbc c21a9013 and
the original 100 Hz CAN FD cadence. C0 samples model y at 7 m;
C1 is max(7 m, speed times 1 s) times selected desired curvature.
This removes the later model-orientation, forecast, damping and
cadence experiments. Model selection/bundles and Panda safety are
unchanged. Restoring the prior behavior does not establish the
physical cause of the reported wobble.
Validation: exact original tree and submodule match; 494 tests and
9146 subtests passed, 178 inapplicable safety skips; canonical
Sunnylink schema; 100 real-sender messages in 100 control cycles,
with counters/checksums, C2/C3 zero and unchanged Panda TX checks.
Assisted-by: OpenAI Codex
Revert the C1 early-release experiment in 7ca2df481 after reported
centering degradation. Restore the exact tracked tree from 4bd841ecc;
model selection and model bundles are unchanged. A different driving
model remains a possible confound, so this does not assign a physical
root cause to either model or controller.
Validation: exact baseline tree match, 267 tests and 3 subtests passed,
and Sunnylink generated schema matches its source.
Assisted-by: OpenAI Codex
Bound the model heading request toward zero using terminal spatial
curvature at the existing preview point. Preserve C0, command limits,
slew, 20 Hz cadence, and the existing Sunnylink selection.
This changes requests before some turn peaks as well as during unwind.
Document that authority tradeoff and distinguish command-level evidence
from unverified physical PSCM release.
Validation: Ford controller/sender/safety tests, Sunnylink schema tests,
independent geometry and replay tests, 28,138 stress CAN round trips,
and 35,775 recorded cycles with 71,550 baseline/candidate CAN checks.
Assisted-by: OpenAI Codex
Compare v7 and recorded v6 unwind instructions on a5 using both common request levels and own-peak thresholds. Preserve the later C0 release and earlier C1 release in the report without claiming a physical improvement.
Exercise saturated release through the actual sender for both signs and every send phase. Require original model clocks for current replay extracts and reject current candidates in the historical unchanged-C1 damping replay. Record the complete 585-test suite and refreshed stress evidence.
Assisted-by: OpenAI Codex
Use the model-predicted one-second station with the existing seven-metre minimum and endpoint hold for both C0 and C1. Remove the extra yaw forecast and scalar-curvature heading reconstruction; retain C2/C3 zero, independent limits and slew, the existing toggle, and 20Hz sends. Reject scalar-only maneuver references explicitly.
Validate model clocks, geometry, CAN delivery, Panda TX acceptance, and randomized boundaries offline. Recorded-input replay does not establish improved physical tracking.
Assisted-by: OpenAI Codex
Snapshot the existing Sunnylink toggle into CarParamsSP so controlsd and the CAN sender agree. Preserve the v6 C0/C1 calculation at 100Hz and Panda safety unchanged. Add actual sender timing, packing, selection, and safety checks; record the a5 frozen-input replay. Physical tracking improvement remains unverified.
Assisted-by: OpenAI Codex
Check both motion publication and embedded filter-state ages before using
calibrated yaw. Rebuild the pose when calibration changes between motion
samples. Four regression cases demonstrate stale, future, missing-state
fallback and calibration-only pose refresh.
Keep the Sunnylink YAML source synchronized with its generated schema and
update the drive guide and source-bound validation record for v6.
Validation: 427 tests and 26 subtests pass; type, lint and settings compiler
checks pass. All 340,757 recorded cycles still match the reviewed candidate,
with 681,514 CAN round trips. Physical steering improvement remains unverified.
Assisted-by: OpenAI Codex
Remove excess-yaw C0 attenuation while retaining full path prediction,
command limits, slew and input-health gates. Valid measured yaw no longer
changes path demand. Update diagnostics and Sunnylink help for v5.
Validate with 356 tests and 26 subtests, 100% controller coverage,
280,636 recorded route cycles and 779,410 Float32/CAN round trips.
These are command checks; physical tracking improvement is not established.
Assisted-by: OpenAI Codex
Remove the hand-chosen 0.15 m / 25% cap on the prediction adjustment.
Retain the available-horizon bound, nonfinite fallback, total field limits,
yaw damping, slew, input gates, two states and zero C2/C3.
Cover full predictions and geometric countersteering in regression tests.
Validate 374 tests plus 26 subtests, four-route replay and 817,346
Float32/CAN round trips. Document command changes without inferring
physical tracking performance from PSCM limits or fixed-input replay.
Assisted-by: OpenAI Codex
Predict nearby C0 from the current model path and selected curvature over
150 ms, bounded to 0.15 m and 25% of the original offset. Retain excess-yaw
damping, two slew states, C1 behavior, input gates, and zero C2/C3 under the
existing default-off Sunnylink toggle.
Validate 372 tests plus 26 subtests, four-route replay, and 817,346
Float32/CAN round trips. Record earlier command-level crossings and the
right-exit damping tradeoff without claiming improved vehicle tracking.
Assisted-by: OpenAI Codex
Select the new controller only on the CAN FD Lightning through a default-off startup toggle. Retire v8 and its setting; disabling restores the original controller or selected observer. Preserve packing, input gates and zero C2/C3.
Fix the existing Params filtered-key buffer lifetime exposed by Sunnylink backup tests. Record 284 tests, 26 subtests and 485238 offline packing round trips; physical calibration remains unapproved.
Assisted-by: OpenAI Codex
Add a two-state C0/C1 core and a separate freshness/timing adapter compatible
with the existing controlsd call. Preserve reviewed endpoint holding, reject
malformed inputs, and keep production selection and safety unchanged.
Validate actual selection/limiting/publication/CAN integration, exact core
replay across 133,550 route cycles, 200,000 randomized and mirrored cycles,
field boundaries, resets and release uncertainty. Record the completed
264-test Ford suite, mutation probes and dependency/source provenance.
Physical tracking remains unvalidated; calibration_approved=false.
Assisted-by: OpenAI Codex
Prevent same-direction C0/C1 growth when measured turning exceeds current and delayed requests during release, retaining request history across driver feedback resets. Permit bounded C1 correction after opposing bias reaches zero when both requests remain undertracked and measured curvature is no longer catching up.
Keep model allocation, gain, field and slew limits, platform selection, and zero C2/C3 unchanged. Add anonymous recorded-input regressions and diagnostics. Validation: 139 Ford tests plus 150 subtests, 46 Sunnylink tests, Ruff, generated settings check, and recorded-command replays. Physical response and stability remain unverified.
Allow release recovery only when fresh measured yaw undertracks both aligned current and delayed requests and PSCM limit is below 2. Unwind the opposing bias toward zero using current yaw error and existing antiwindup; preserve C0, base geometry, gains, rates, and safety guards.
Add mirrored unit checks and a sanitized recorded turn-exit regression. Validate with 142 tests and 97 subtests, full-route frozen-input replay, large-turn retention, CAN packing, diagnostics, and generated Sunnylink schema checks. Physical improvement remains unvalidated.
Reuse the existing model-pose allocator for aligned large maneuvers while encoding remaining selected curvature as C0/C1 and keeping C2/C3 zero. Permit measured heading backoff during release or PSCM limits without turning model-base changes into stored bias.
Validate with 127 tests and 67 subtests, including recorded large-turn retention, release and reversal, repeated-measurement backoff, CAN packing, logging, and Sunnylink schema checks. Replay checks command behavior; enabled vehicle tracking remains unvalidated.
Retain the absolute desired-curvature base and unchanged C0, while adding
measured yaw-error correction to C1 under fresh PSCM status. Preserve command
limits, reset on override or unusable status, and release stored correction
with the base request. Admit reachable partial increments at host slew limits.
Publish PSCM enums with original CAN receipt timestamps through carStateSP.
Add telemetry, status/driver guards, CAN roundtrip tests and recorded fixtures.
Validation: focused suite 116 tests and 38 subtests; Ruff, settings compilation
and diff checks pass. Production replay covers 52,273 route80 cycles; no-status
fallback preserves v4 over 246,961 cycles / 43 segments. Physical stability and
the reported 85-degree plateau remain unvalidated; EPS limits can inhibit the
new correction.
Derive full absolute C1 heading from the same selected curvature as C0, retaining existing bounds and independent slew. Keep the former filtered model heading as a diagnostic comparison and preserve input validity gates.
Add real route80 command regressions and release/reversal checks. All 97 focused tests pass; 43-segment replay preserves C0 and gates exactly and matches the independent C1 candidate. Physical tracking and stability remain unvalidated for this revision.
Encode the selected bounded curvature as C0 with an 8 m minimum preview while retaining full model-heading C1. Limit each channel independently so C1 transitions cannot delay C0 release, and validate the selected action source timestamp.
Add action, release, source-freshness, CAN and route regressions. Document the slow-turn reference disagreement and the limits of frozen-motion replay; physical centering remains unvalidated.
Replace the weak nominal acceleration conversion with C2-free spatial path requests. Align retained model geometry using measured CAN yaw before filtering model innovations, and preserve large-turn demand and straight-path centering.
Validate with seven recorded maneuver episodes, full-route command replay, real CAN packing and focused controller/settings tests. Physical closed-loop behavior remains unvalidated.
Track the bounded planner reference through C0 and delay-aware PI/rate feedback through C1. Remove the failed Shared Path toggle and add a default-off, Lightning RL38-specific Virtual Angle setting. Reject stale inputs and disable outgoing lateral requests when the path is invalid.
Validation: 85 focused tests plus 22 subtests, native Params, real CAN packing, settings generation and Ruff passed. Frozen route78 replay attenuates the observed command forcing; physical stability and turn authority remain unvalidated.
Separate holding demand, bounded pose feedback, and nominal coefficient allocation. Add a default-off Sunnylink selector with startup diagnostics and preserve the existing controller when disabled.
Assisted-by: Codex
Isaac Barham