DiePre Embodied Bridge

DiePre 具身桥接层 —— 将2D视觉检测桥接到3D空间理解和机器人动作规划，vision-action-evolution-loop 的具体实现

Security Scan

The skill's documentation describes a non-trivial Python toolchain that reads/writes evolution logs and exposes a DiePreEmbodiedBridge API, but the package contains only docs (no implementation or install spec), which is incoherent and deserves caution before use.

Purpose & Capability

SKILL.md describes a Python class, six callable tools (diepre_vision.analyze, dimension_estimator, etc.), and examples that import skills.diepre_embodied_bridge, yet the registry entry contains no code files or install spec to provide those implementations. The stated purpose (2D→3D bridge and robot planning) would legitimately require code, binaries, or dependencies, none of which are present in this package—this mismatch is a functional/integrity concern.

Instruction Scope

The instructions explicitly direct the agent to write and read local files (evolution_log/{task_id}.json, params/evolved_params.json) and to run periodic heartbeat tasks that aggregate failures and update parameters. Those behaviors are coherent for an evolving robotics skill, but they introduce persistent local state and periodic background processing; verify you are comfortable with automatic logfile writes and periodic processing in the chosen workspace. The instructions do not request unrelated system credentials or specify external network exfiltration endpoints.

Install Mechanism

There is no install specification and no implementation files—only documentation. The README suggests 'clawhub install' or copying a skills/ directory, but the package does not include the code to be installed. This creates ambiguity: either the skill is a documentation stub (harmless but non-functional), or it expects the installer to fetch code from the referenced GitHub homepage (which is external and not vetted here). Lack of an explicit, auditable install mechanism is a risk.

Credentials

The skill declares no required environment variables, credentials, or config paths, which is proportionate to the documentation-only package. The described runtime might in practice need model weights or hardware access (GPU drivers) but none are requested here—confirm what external resources the real implementation (if obtained) requires before running.

Persistence & Privilege

The skill's design includes persistent logs and a heartbeat-driven evolution loop (HEARTBEAT.md) that periodically reads logs and writes evolved parameters. It does not request elevated privileges or global 'always' inclusion, but the persistent file writes could accumulate data over time; if you install a real implementation, consider where logs are stored and retention/permission policies.

What to consider before installing

This package currently contains only documentation describing a substantial Python toolchain but provides no implementation or install steps that fetch code. Before installing or running anything: 1) verify the referenced GitHub repository actually contains the implementation and review its code for network calls, subprocess execution, or any secrets handling; 2) do not run unreviewed installation scripts or copy code from untrusted sources — prefer cloning the repository and auditing it first; 3) be aware the skill will create and update local logs (evolution_log/ and params/evolved_params.json); ensure those files won't leak sensitive data and run the skill in a sandboxed environment if possible; 4) confirm what external model weights, binaries, or drivers the real implementation needs (they may require extra permissions or downloads); and 5) if you cannot obtain or audit the implementation, consider this package non-functional and avoid installing it.

DiePre Embodied Bridge Skill

元数据

字段	值
名称	diepre-embodied-bridge
版本	1.0.0
作者	KingOfZhao
发布日期	2026-03-31
置信度	96%

核心哲学

vision-action-evolution-loop 定义了抽象的五阶段闭环。 本 Skill 是它的具体实现层——聚焦于"如何把2D线条变成3D动作"。

认知节点关系：

vision-action-evolution-loop (父: 抽象闭环)
    └── diepre-embodied-bridge (本Skill: 具体实现)
            ├── diepre-vision-cognition (上游: 2D检测)
            └── diepre-action-memory (下游: 动作记忆, 未来)

三大核心认知

1. 已知几何估算（非通用3D重建）

包装盒不是复杂场景，是已知几何体。不需要 NeRF / Gaussian Splatting / SfM。

输入: 2D DXF + FEFCO类型 + 纸板厚度
算法: FEFCO规则引擎 + 2D尺寸 → 3D展开坐标 → 折叠矩阵
输出: 三维空间坐标 (x,y,z) + 折叠顺序 + 面法向量
硬件: M1 Max 轻松运行（纯CPU计算，<100ms）

为什么排除 NeRF/Gaussian Splatting？

• 包装盒是平面折叠结构，不是复杂3D场景
• NeRF需要数百张照片+GPU集群训练，M1 Max跑不动
• Gaussian Splatting需要密集视角，生产环境不现实
• 已知几何估算：1张照片+FEFCO规则→3D，秒级完成

2. MCP 工具链（Tool-Augmented）

OpenCV 管道封装为可调用工具，VLA 模型调用工具而非处理原始图像：

tools = {
    "detect_dieline": {
        "input": "image_path: str",
        "output": "dxf_path: str, confidence: float",
        "impl": "diepre_vision.analyze"
    },
    "estimate_dimensions": {
        "input": "dxf_path: str",
        "output": "length, width, height, thickness_mm",
        "impl": "dimension_estimator.from_dxf"
    },
    "identify_fefco_type": {
        "input": "dxf_path: str, layout_features: dict",
        "output": "fefco_type: str (e.g. 0201, 0427)",
        "impl": "fefco_classifier.classify"
    },
    "calculate_fold_sequence": {
        "input": "fefco_type: str, dimensions: dict, material: str",
        "output": "ordered_steps: list[FoldStep]",
        "impl": "fold_planner.plan"
    },
    "compute_grasp_points": {
        "input": "fold_sequence: list[FoldStep], material_thickness: float",
        "output": "grasp_points: list[GraspPoint] (xyz + force + angle)",
        "impl": "grasp_calculator.compute"
    },
    "estimate_quality": {
        "input": "image_path: str, expected_dimensions: dict",
        "output": "quality_score: float, defects: list",
        "impl": "quality_checker.evaluate"
    }
}

3. 自迭代进化机制

执行任务 → 记录结果 → 提取失败模式 → 调整参数 → 下次优化

具体流程:
1. 每次任务执行完，写入 evolution_log/{task_id}.json:
   {
     "task_id": "diepre_20260331_001",
     "input": {"image": "...", "fefco": "0201", "material": "B flute"},
     "execution": {"steps": [...], "timing_ms": 3400},
     "result": {"success": false, "fail_step": 3, "error": "grasp_slip"},
     "params_used": {"grasp_force": 2.5, "approach_angle": 45}
   }

2. 定期扫描 evolution_log/，提取失败模式:
   - grasp_slip 在 B flute 上发生频率 73% → 提高抓取力
   - fold_sequence 错误在 FEFCO 0427 上频率 40% → 修正折叠规则

3. 更新参数文件 params/evolved_params.json:
   {"B_flute_grasp_force": 3.2, "0427_fold_override": [...]}

4. 下次任务加载 evolved_params.json，用优化后参数执行

安装命令

clawhub install diepre-embodied-bridge
# 或手动安装
cp -r skills/diepre-embodied-bridge ~/.openclaw/skills/

调用方式

from skills.diepre_embodied_bridge import DiePreEmbodiedBridge

bridge = DiePreEmbodiedBridge(workspace=".")

# 单次执行
result = bridge.execute(
    image_path="path/to/box_photo.jpg",
    material="B flute",
    thickness_mm=3.0
)

print(result.fefco_type)          # "0201"
print(result.dimensions)          # {"L": 300, "W": 200, "H": 100}
print(result.fold_sequence)       # [FoldStep(...), ...]
print(result.grasp_points)        # [GraspPoint(x=150,y=0,z=50,force=3.2), ...]
print(result.quality_score)       # 0.92
print(result.confidence)          # 0.96

# 自迭代: 注入失败反馈
bridge.record_failure(
    task_id="diepre_20260331_001",
    fail_step=3,
    error_type="grasp_slip",
    context={"material": "B flute", "grasp_force": 2.5}
)

# 查看进化状态
stats = bridge.evolution_stats()
print(stats.total_tasks)          # 47
print(stats.failure_rate)         # 0.12
print(stats.top_failure_modes)    # [("grasp_slip", 8), ("fold_error", 4)]

学术参考文献

1. From 2D CAD to 3D Parametric via VLM — 2D→3D桥接，参数化建模
2. Tool-Augmented VLLMs as Generic CAD Task Solvers (ICCV 2025) — 工具增强策略，MCP工具链的理论基础
3. Vlaser: Synergistic Embodied Reasoning — 抓取点计算+力控参数
4. Efficient VLA Models — 本地部署优化（M1 Max适用）
5. SAGE: Multi-Agent Self-Evolution — 自迭代进化的学术对应
6. Self-evolving Embodied AI — 记忆自更新+参数进化