87 lines
6.4 KiB
Markdown
87 lines
6.4 KiB
Markdown
# New Window Prompt: EM Longitudinal Rolling Planning — Phase 3
|
|
|
|
Work in `D:\Users\Desktop\项目\prakrobot\ParkingRobot`. This is an implementation task, not a design discussion. Read these files completely before editing:
|
|
|
|
- `docs/superpowers/specs/2026-08-05-em-longitudinal-rolling-planning-design.md`
|
|
- `docs/superpowers/plans/2026-08-05-em-longitudinal-rolling-planning.md`
|
|
- `docs/superpowers/plans/2026-08-05-em-longitudinal-rolling-four-phase-roadmap.md`
|
|
- `docs/superpowers/handoffs/2026-08-05-em-longitudinal-rolling-phase-2.md`
|
|
|
|
Phase 2 code ends at `59d13e5`; Task 3 and Task 4 also have local fixups `dbc7b7c` and `26bd822` because the dirty shared worktree prevented autosquash. Preserve all actual interfaces listed in the handoff, especially `EmLongitudinalMode`, `LongitudinalTerminalSchedule`, mode-explicit `LongitudinalPlanningInput`, `PathUpperBoundS`, `StopBoundaryPathS`, and `PathSpeedLimit.HasStopBoundary`.
|
|
|
|
Execute only original-plan Tasks 6 and 7. Do not begin Task 8 or later. Use `executing-plans`, TDD for every behavior change, `systematic-debugging` for any unexpected build/test failure, and `verification-before-completion` before every commit and before phase handoff. No brainstorming is needed.
|
|
|
|
Start by recording:
|
|
|
|
```powershell
|
|
git branch --show-current
|
|
git status --short
|
|
git log -10 --oneline
|
|
git show --stat --oneline 59d13e5
|
|
```
|
|
|
|
The branch is `trajplanner` and the worktree has substantial unrelated user changes. Never reset, checkout, clean, broadly stage, stash user files, use `git add .`, or use `git add -A`. Stage only the listed task files and run `git diff --cached --name-only` before every commit.
|
|
|
|
## Task 6 — mode-aware trajectory publication and validation
|
|
|
|
Files:
|
|
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Contracts/EmTrajectoryMetadata.cs`
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Trajectory/TrajectorySampleSchedule.cs`
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Trajectory/EmTrajectoryAssembler.cs`
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Validation/EmTrajectoryValidator.cs`
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Facade/EmPlanningService.cs`
|
|
- `ClumsyPilot/tests/EMPlannerVerificationHost/TrajectoryChecks.cs`
|
|
- `ClumsyPilot/tests/EMPlannerVerificationHost/TrajectoryObservationChecks.cs`
|
|
- `ClumsyPilot/tests/EMPlannerVerificationHost/ExecutorChecks.cs`
|
|
- `ClumsyPilot/tests/EMPlannerVerificationHost/CoordinatorChecks.cs`
|
|
|
|
Write and observe failing trajectory tests first. Add `EmLongitudinalMode` to trajectory metadata and pass the mode from `EmPlanningService`. For Rolling and Approach, sample exactly the ST candidate knots: no synthetic zero-speed hold, no terminal boundary anchor, and a nonzero terminal state remains valid when ordinary limits are met.
|
|
|
|
For Exact mode, verify the complete internal static tail from `LongitudinalTerminalSchedule.GetStabilizationStartIndex`: every tail knot must hold the same PathS with `U=0`, `A=0`, and zero tail jerk. Mark only the first tail knot as the real Goal/GearSwitch boundary anchor; then append `ZeroSpeedHoldSeconds` only after the QP horizon, without duplicate timestamps. Add `TerminalAccelerationNotZero` to publication failures and make `EmTrajectoryValidator` always retain ordinary finite, kinematic, jerk, and collision checks while applying terminal-anchor/tail checks only to Exact mode. Do not weaken any threshold or add chassis behavior.
|
|
|
|
Required checks before the Task 6 commit:
|
|
|
|
```powershell
|
|
dotnet run --project ClumsyPilot/tests/EMPlannerVerificationHost/EMPlannerVerificationHost.csproj -- trajectory
|
|
dotnet run --project ClumsyPilot/tests/EMPlannerVerificationHost/EMPlannerVerificationHost.csproj -- coordinator
|
|
dotnet run --project ClumsyPilot/tests/EMPlannerVerificationHost/EMPlannerVerificationHost.csproj -- executor
|
|
dotnet run --project ClumsyPilot/tests/EMPlannerVerificationHost/EMPlannerVerificationHost.csproj -- trajectory-observation
|
|
```
|
|
|
|
Commit only the Task 6 files with:
|
|
|
|
```powershell
|
|
git commit -m "feat: publish rolling trajectories without stop tails"
|
|
```
|
|
|
|
## Task 7 — previous-trajectory longitudinal soft reference
|
|
|
|
Files:
|
|
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Longitudinal/LongitudinalPreviousTrajectorySeedBuilder.cs` (new)
|
|
- `ClumsyPilot/ParkrobTrajplanner/EMPlanner/Facade/EmPlanningService.cs`
|
|
- `ClumsyPilot/tests/EMPlannerVerificationHost/LongitudinalModelChecks.cs`
|
|
- `ClumsyPilot/tests/EMPlannerVerificationHost/EmPlanningServiceChecks.cs`
|
|
|
|
Write and observe failing tests first. Create `LongitudinalPreviousTrajectorySeed` with `PathS`, `ProgressSpeedMetersPerSecond`, and `Empty`. Its builder must resample the previous `EmTrajectory` at `newEffectiveAtUtc + knotTime`, linearly interpolate world position and signed speed, project monotonically onto the current `LateralPath`, and publish absolute progress speed. Any incompatible direction, segment, missing trajectory, or out-of-range sample must return `Empty` without throwing into planning.
|
|
|
|
In `EmPlanningService`, construct current ST knot times after lateral planning, build this seed from the previous trajectory/current path/effective time/segment/direction, and pass the two reference arrays into the existing mode-explicit `LongitudinalPlanningInput`. Confirm the objective receives nonzero previous-S and previous-U soft-reference terms only for a valid seed; these remain soft terms, never terminal constraints. Keep LS slicing on `WindowEndReferenceS`.
|
|
|
|
Required checks before the Task 7 commit:
|
|
|
|
```powershell
|
|
dotnet run --project ClumsyPilot/tests/EMPlannerVerificationHost/EMPlannerVerificationHost.csproj -- longitudinal-model
|
|
dotnet run --project ClumsyPilot/tests/EMPlannerVerificationHost/EMPlannerVerificationHost.csproj -- em-planning-service
|
|
```
|
|
|
|
Commit only the Task 7 files with:
|
|
|
|
```powershell
|
|
git commit -m "feat: reuse prior trajectory in longitudinal planning"
|
|
```
|
|
|
|
Global semantics are non-negotiable: `DistanceHorizonMeters` is LS look-ahead only; `TimeHorizonSeconds` is one ST horizon only; only Goal/GearSwitchApproach may require Exact zero speed; `ZeroSpeedHoldSeconds` is outside the QP horizon; do not loosen jerk, acceleration, solver, or validation thresholds; and retain `OBSERVE_ONLY` with no chassis command.
|
|
|
|
At phase end, rerun `trajectory`, `em-planning-service`, `coordinator`, and `executor`; run `git diff --check`, `git log -5 --oneline`, and `git status --short`. Use `writing-plans` to create the phase-3 handoff and phase-4 prompt, commit only those two documents, and stop without implementing Task 8.
|