🗺️ DDS 节点感知地图

xspace 21节点·14技能·21Topic · web可感知·可定义·可应用 · 2026-07-13

Skill 0 · 传感器采集

节点: realsense_rgb / realsense_depth / force_sensor / tactile_array
Topic: sensor_rgb / sensor_depth / sensor_force_6d / sensor_tactile
API:  gds.publish("sensor_force_6d", [fx,fy,fz,tx,ty,tz])
Web:  → VLA-T训练输入 / 数据质量L0校验
模型: Sys21 VLA-T 触觉编码器

Skill 1 · 关节状态

节点: joint_state_raw / joint_calib
Topic: joint_pos_7d / joint_vel_7d / joint_torque_7d
API:  gds.get("joint_pos_7d") → [j1..j7]
Web:  → Orin管线实时监控 / L1格式校验
模型: Sys1 ACT 状态编码器

Skill 2 · ACT推理

节点: infer_sys1_act
Topic: infer_sys1_act / act_action_chunk
API:  gds.get("infer_sys1_act") → (50,6) tensor
Web:  → 远程控制面板引擎状态 / L2清洗
模型: Sys1 ACT 52M (xspace 4060)

Skill 3 · SmolVLA推理

节点: infer_sys11_smolvla
Topic: infer_sys11_smolvla / smolvla_output
API:  gds.publish("infer_sys11_smolvla", img, lang)
Web:  → 系统接口页引擎切换 / L2校验
模型: Sys11 SmolVLA 450M (xspace→web)

Skill 4 · VLA-T力控

节点: infer_sys21_vtla / lissajous_search
Topic: infer_sys21_vtla / vtla_force_feedback
API:  gds.get("infer_sys21_vtla") → (6轴力+置信度)
Web:  → Orin管线Lissajous参数 / L3归一化
模型: Sys21 VLA-T ~470M (web 4090)

Skill 5 · GR00T决策

节点: infer_sys22_groot / groot_planner
Topic: infer_sys22_groot / groot_action_40step
API:  gds.publish("infer_sys22_groot", obs) → eef+joint+grip
Web:  → 数据闭环面板 / L3增强
模型: Sys22 GR00T 7B (web 4090)

Skill 6 · AOI检测

节点: aoi_capture / aoi_analyze
Topic: aoi_capture / aoi_result / aoi_detections
API:  gds.get("aoi_result") → {code, msg, detections[]}
Web:  → Orin管线AOI×6 / L2清洗
模型: 4060:10082 HTTP POST

Skill 7 · 权重同步

节点: weight_sync / version_check
Topic: weight_vtla / weight_groot / model_version
API:  gds.publish("weight_vtla", path, hash)
Web:  → 数据闭环ECS中转 / L4入仓
模型: 全模型 版本管理

Skill 8-13 · 扩展技能

Skill 8:  data_logger      → 所有数据持久化存储
Skill 9:  data_quality      → L0-L4五层质量检查
Skill 10: sim_bridge        → Mac旁路仿真桥接
Skill 11: orin_bridge       → Orin真机部署桥接
Skill 12: train_trigger     → 自动触发4090训练
Skill 13: deploy_service    → 权重→Orin自动部署
API:  gds.to_json("all")    → 全部数据JSON给web

DDS API 速查

from dds_skill import gds
gds.publish("sensor_force_6d", [1.2, 0.3, 5.0, 0, 0, 0])  # 发布
gds.get("infer_sys1_act")                        # 读取最新值
gds.subscribe("aoi_result", callback_fn)          # 订阅
gds.to_json("all")                                # 全部JSON→Web
gds.quality_check()                               # 运行质量检查