PROJECT 01 / SMARTCAM
从深度感知到机器人控制Connecting depth perception to robot control
围绕 RGB-D 感知、边缘视觉与控制接口，探索可调试、可追溯的机器人感知链路。Exploring a debuggable, traceable sensing pipeline that connects RGB-D perception, edge vision and robot control.
PERCEPTION → ACTIONSYS / SMARTCAM
CONCEPTUAL SYSTEM DIAGRAM原创示意 / 非实物照片
01 / PROBLEM
要解决什么问题The engineering problem
机器人需要把相机输出转化为控制侧可使用的信息。工程重点不仅是目标检测，还包括深度流、时间关系、日志与系统联调。A robot needs camera outputs in a form its controller can use. The engineering work spans object detection, depth streams, timing, logging and integration.
02 / ARCHITECTURE
系统如何连接How the system connects
- 01RealSense：彩色与深度输入RealSense: color and depth input
- 02边缘视觉：YOLOv8n 与感知处理Edge vision: YOLOv8n and perception processing
- 03ROS 2 / 控制接口：消息与系统联调ROS 2 / control interfaces: messages and integration
- 04日志与测试：回看行为及诊断问题Logs and tests: inspect behavior and diagnose issues
以上为概念级结构梳理，不是实物接线图或验证报告。A conceptual architecture summary, not an as-built wiring diagram or validation report.
03 / CONTRIBUTIONS
我的工作与参与My work & contributions
- 编写相机采集、感知流与机器人接口相关脚本。Develop camera acquisition, perception pipeline and robot interface scripts.
- 围绕 RGB-D 输入与边缘设备开展集成和联调。Integrate RGB-D inputs with edge computing hardware.
- 通过测试日志和训练记录支持问题定位与迭代。Use test logs and training records to support diagnosis and iteration.
04 / VALIDATION & LIMITS
有证据的结论，明确的边界Evidence, with clear boundaries
现有资料包含相机脚本、检测训练配置、结果与测试日志。端到端制动时延、带宽收益和实机安全性能尚不作为已验证结论展示。VLA 是后续研究扩展。Available materials include camera scripts, detection training configurations, results and test logs. End-to-end braking latency, bandwidth savings and operational safety are not presented as verified results. VLA remains a research extension.
05 / MEDIA
工程媒体与后续补充Engineering media
真实工程媒体 · 待审核补充Engineering media · pending review
本页展示原创系统示意图。实际机器人照片与演示视频将在完成来源及隐私审核后补充。This page uses an original conceptual system diagram. Robot photographs and demonstration videos can be added after provenance and privacy review.
NEXT PROJECT