ROS TeleOp

I built a small four-wheeled differential drive robot controlled via ROS teleoperation, designed to explore the integration of computer vision (OpenCV) into mobile robotics. The bot combined a Raspberry Pi as the onboard computer with an Arduino Uno for low-level motor control.

System Overview:
The control pipeline consisted of a ROS master running on the Raspberry Pi, where teleoperation nodes published velocity commands (/cmd_vel). These were intended to be sent over serial to the Arduino Uno, which interfaced with the L298N motor driver to actuate the DC motors. Simultaneously, the Raspberry Pi Camera provided a video feed into ROS topics for OpenCV-based processing. The power system was managed by a 3S 2200 mAh LiPo battery with a power distribution board, supplying regulated power to both the Pi and Arduino.

Key Features & Implementation:

• Mechanical Platform: Four-wheel drive base with DC motors driven via an L298N motor driver shield.

• Electronics & Power:

  • Arduino Uno for motor actuation.

  • Raspberry Pi 4 running Ubuntu as the ROS host.

  • Powered by a 3S 2200 mAh LiPo battery with a dedicated power distribution board and voltage regulators.

• Control Architecture:

  • Arduino Uno controlled motors via L298N.

  • Serial communication intended between Raspberry Pi (ROS Master) and Arduino.

  • Teleop using ROS keyboard teleop node publishing to /cmd_vel.

• Perception: Raspberry Pi Camera for computer vision tasks, intended for OpenCV-based object tracking and later integration into ROS topics.

Challenges & Issues Faced:

• Camera–ROS Compatibility:

  • Pi Camera support was limited in ROS1, incompatible with standard drivers (cv_camera, usb_cam).

  • Switching to ROS2 (Foxy) enabled camera streaming and OpenCV integration.

• Motor Control in ROS2:

  • ROS2 lacked rosserial_arduino, preventing direct Arduino motor control.

  • This forced a choice: ROS1 for motor actuation but no camera, or ROS2 for camera but no Arduino serial.

Possible Fixes That I Couldn't Implement:

• Use the rosserial_python fork for ROS2 or migrate to micro-ROS for embedded ROS2 integration.

• Replace Pi Camera with a USB webcam for ROS1 compatibility.

• Bypass rosserial entirely: write a custom Python-serial ROS node on the Pi to interface with Arduino.

• Use ros1_bridge to connect ROS1 (Arduino) with ROS2 (Camera/Vision).

Why This Project Matters:
This project introduced me to the complexities of ROS-based ground robotics, teaching me about teleop control, hardware/software integration, and the trade-offs between ROS1 and ROS2 ecosystems. It emphasized the importance of designing modular systems where perception, control, and actuation can evolve independently.

Next Steps Would Be To:

• Transition to micro-ROS on Arduino for ROS2 compatibility and reliable motor control.

• Add wheel encoders + odometry publishing for localization and SLAM.

• Integrate with the ROS2 Navigation Stack (Nav2) using LiDAR or depth sensors.

• Expand vision capabilities (object following, line detection) with OpenCV and the Pi Camera.

• Evolve toward a fully autonomous differential drive robot capable of indoor exploration and mapping.