feat(tasks): add declarative physical handover

[yuecideng]

Description

Stack

• Layer: 21/28
• Base: feat/expert-program-registration-runtime-catalog
• Depends on: feat(expert-program): add pre-simulation runtime catalog #504
• Followed by: feat(skills): expose per-expectation effect outcomes #477

Add the Dual-UR5/PGI declarative HandOver vertical slice, its exact pose/effect integration, and real contact-dynamics acceptance coverage. The runtime guards held objects in flight and detects physical loss instead of hiding it with a synthetic attachment.

The layer also fixes solver configuration construction so finalized parameters reach subclasses correctly. HandOver completes through ordinary commands, physical evidence, settling, and validation without simulator-side object repair.

Refs #471
Refs #474

Type of change

• New feature (non-breaking change which adds functionality)

Screenshots

Not applicable.

Validation

• Focused coverage: tests/gym/envs/tasks/test_hand_over.py, simulation-handover/policy tests, atomic runner/session tests, and solver tests
• Final affected-suite regression on the stack tip: 1215 passed, 2 skipped, 8 deselected
• Changed Python files pass Black 26.3.1; the Sphinx build and rollout-report drift check pass at the stack tip

Checklist

• Changed Python files pass Black 26.3.1.
• Corresponding public/design documentation is included in this stack.
• Tests cover the affected behavior.
• No dependency update is required.

[greptile-apps]

Greptile Summary

The PR adds a declarative Dual-UR5 handover task and introduces phase-aware physical held-object guards with row-local state reconciliation. It also aligns simulation runner cadence, preserves position-controller preload during settling, and fixes solver configuration construction.

• Adds the packaged handover environment, robot/scene profile, semantic poses, configuration, and physical-simulation acceptance coverage.
• Observes held-object loss during named trajectory phases and routes failed rows into bounded retry or external recovery.
• Adds configurable continuation behavior during effect verification and completion.
• Constructs concrete solver configurations with finalized constructor parameters before subclass post-initialization.

Confidence Score: 5/5

The PR appears safe to merge, with no concrete blocking or independently actionable non-blocking issue identified.

The new guard path correlates fresh evidence to the active attempt and waypoint, preserves healthy batch rows during reconciliation, and enforces timeout recovery before further command dispatch; the solver and handover configuration paths are likewise consistent with their callers and tests.

Important Files Changed

Filename	Overview
embodichain/lab/sim/atomic_actions/execution.py	Adds correlated phase-aware guard requests/results, row-local relation invalidation, and retry or external-recovery transitions.
embodichain/lab/sim/atomic_actions/runner.py	Invokes held-object guards before due commands and supports suppressing observed-position holds during effect verification.
embodichain/lab/sim/skills/compiler.py	Grounds Pick, Place, and HandOver held-object invariants against named trajectory segments and physical effect monitors.
embodichain/lab/sim/skills/runtime.py	Collects phase-scoped evidence, computes loss masks, and produces authorized removal-only state reconciliation.
embodichain/lab/gym/envs/expert_program/simulation_policies.py	Preserves drive targets for active settling rows while holding inactive rows at measured positions.
embodichain/lab/sim/solvers/base_solver.py	Passes recognized concrete configuration fields through construction so subclass post-initialization observes finalized values.
embodichain_tasks/embodichain_tasks/tableware/hand_over.py	Defines the declarative Dual-UR5/PGI handover environment, calibrated physical scene, profile, and registration.
embodichain_tasks/configs/expert_program/tableware/hand_over.yaml	Declares the Pick-to-HandOver workflow, settling policy, and delivery-pose validator.
embodichain_tasks/configs/gym/hand_over/dual_ur5.json	Adds the runnable task configuration with dual-arm geometry, controller parameters, and contact dynamics.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Declarative Pick / HandOver call] --> B[Compiler grounds motion and effect guards]
    B --> C[ExecutionRunner obtains fresh observation]
    C --> D{Held-object guard passes?}
    D -- Yes --> E[Dispatch next command]
    D -- Loss with retry --> F[Invalidate failed-row relation]
    F --> G[Replan within retry budget]
    D -- Loss without retry --> H[Emit recovery required]
    E --> I[Terminal physical effect verification]
    I --> J[Commit verified state]
    J --> K[Wait-stable using live target qpos]
    K --> L[Validate delivered object pose]

Loading

_{Reviews (1): Last reviewed commit: "feat(atomic-actions): guard in-flight he..." | Re-trigger Greptile}

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[yuecideng]

Folded into #480 during stacked-PR consolidation. Its commits remain included in #480; the remote branch is retained for traceability.