TS-ROS2 Robotics Platform

TS-ROS2 is a modular ROS2-based robotics system for waypoint navigation, perception, and manipulation.

It evolved from a single navigation stack into a layered robotics platform with multiple versioned system generations.

System Versions

v1.0 — Legacy Stable System

• Original UDP protocol (v1)
• Waypoint-based navigation system
• Basic P-controller robot control loop
• Monolithic ROS2 workspace structure
• Android app compatibility (v1 protocol)

v1.0 is stable and frozen (no further feature development)

v2.0 — Next-Generation Platform (Active Development)

• Breaking change: UDP protocol v2
• Modular ROS2 architecture redesign
• Added perception pipeline (CV system via Docker runtime)
• Added manipulation subsystem
• Expanded Android UI modules
• Multi-layer system design:
  • Core navigation layer
  • Perception layer
  • Manipulation layer
  • Communication layer

v2.0 is the active development line of TS-ROS2

Compatibility

Feature	v1 App	v2 Stack
UDP v1 protocol	✔	❌
UDP v2 protocol	❌	✔
Navigation system	✔	✔
CV pipeline	❌	✔
Manipulation	❌	✔

System Overview

TS-ROS2 is structured as a layered robotics system:

Core Layer

• Robot motion control
• Odometry processing
• Waypoint execution
• State machine (navigate / pause / return)

Communication Layer

• UDP-based Android ↔ Robot interface
• ROS2 topic bridge
• Real-time command synchronization

Perception Layer (v2)

• Camera input processing
• GPU inference (TensorRT / ONNX fallback)
• Docker-isolated CV runtime
• Shared-memory frame pipeline

Manipulation Layer (v2)

• Gripper control logic
• Detection-based interaction
• Distance-aware decision system

Communication Architecture

v1 (Legacy)

• Simple waypoint batch format
• Fixed schema
• Android app compatibility

v2 (Current)

• Structured frame system
• Extended metadata support
• Not backward compatible with v1

ROS2 Workspace Structure

communication/
control/
navigation/
hardware_interface/
perception/      (v2)
manipulation/    (v2)
robot_interface/
target_setter/

Each module is independently testable and replaceable.

Perception Runtime (v2)

• Docker-based CV execution environment
• GPU acceleration support (CUDA / TensorRT)
• ONNX fallback for portability
• Shared memory bridge with ROS2 inference nodes

Design Philosophy

• Modular robotics architecture over monolithic design
• Clear separation of concerns:
  • control
  • perception
  • communication
• Optional subsystem design (CV / manipulation)
• Deterministic navigation core remains stable across versions

Supported Hardware

• MSI Cyborg 15 (RTX 3050)
• NVIDIA CUDA-capable systems
• Jetson devices (experimental support)
• Standard ROS2-compatible Linux systems

Project State

• v1.0 → legacy stable system (frozen)
• v2.0 → active robotics platform development

custom_messages not found at build

• Make sure the custom_messages package is inside ~/ros2_ws/src and run colcon build again before sourcing.

For technical insight, please read development_notes.md

Known Issues

• Yaw error — heading is off
• Residual position error — P controller leaves steady-state error near the target.
• Waypoint edit bug — edits oapply to the wrong version of planning due to mistake on the app.
• No emergency stop — not yet implemented on the app.
• No launch file — nodes must be started manually.