Architecture¶

This document describes the system architecture of the Robot Environment.

System Overview¶

The Robot Environment is built in several layers to ensure a clear separation of concerns.

graph TD Env[Environment Layer] --> Robot[Robot Control Layer] Env --> Vision[Vision Layer] Env --> Workspace[Workspace Layer] Robot --> RC[RobotController Abstract] RC --> NiryoRC[NiryoRobotController] RC --> WidowXRC[WidowXRobotController] Vision --> FG[FrameGrabber Abstract] FG --> NiryoFG[NiryoFrameGrabber] FG --> WidowXFG[WidowXFrameGrabber] Workspace --> WS[Workspace Abstract] WS --> NiryoWS[NiryoWorkspace]

Components¶

Environment Layer¶

The Environment central orchestrator coordinates all subsystems. It manages the object memory (ObjectMemoryManager) and ensures that camera updates and robot commands are executed in a thread-safe manner.

Robot Control Layer¶

Provides a high-level API for pick-and-place operations. The abstract RobotController class allows support for different hardware backends.

Vision Layer¶

Responsible for capturing images and streaming them via Redis. It integrates with the vision_detect_segment package for AI-powered object detection.

Workspace Layer¶

Manages spatial boundaries and provides transformations between camera coordinates (pixels) and world coordinates (meters).

Data Flow¶

The following diagram shows the data flow during a typical update cycle:

sequenceDiagram participant E as Environment participant F as FrameGrabber participant R as Redis participant V as Vision Service participant M as Memory Manager loop Camera Update E->>F: get_current_frame() F->>R: Publish Image R->>V: Process Image V->>R: Publish Detections R->>E: get_detected_objects() E->>M: update(detections) end