New ROS 2 Features - Mini Pupper 2
Hello everyone,
I’m Kishan Grewal, a Robotics and AI student at University College London (UCL). Over the summer, I interned at MangDang in Hong Kong, where I extended the ROS 2 software stack for the Mini Pupper 2 quadruped robot.
My work broadened the Mini Pupper 2’s ROS 2 ecosystem with additional packages that work across different hardware configurations. Depending on your setup, you can now explore person tracking, autonomous navigation, or multi-robot coordination - all through ROS 2.
Vision-Based Tracking
I developed a tracking node that performs real-time person detection and tracking using a YOLO11n ONNX model with multi-object detection and IMU feedback for stable control.
If you have a Mini Pupper 2 with a compatible Pi camera, this package lets the robot visually follow a person while maintaining orientation. The system includes a lightweight Flask web interface and RViz visualization.
Code - mini_pupper_tracking package
Multi-Robot Fleet Coordination
I created a fleet package that uses ROS 2 namespacing and synchronised state estimation to enable cooperative motion, shared heading control, and distributed communication.
So if you have access to multiple Mini Pupper 2 robots, try out coordinated multi-robot teleoperation.
Code - mini_pupper_fleet package
Navigation and SLAM
If you’ve fitted your Mini Pupper 2 with a LiDAR sensor, the navigation stack now allows you to use SLAM Toolbox and Nav2 to perform reliable map-based localization and path planning. The configuration is currently tuned for table-area navigation but can be altered for other environments.
This setup enables a Mini Pupper 2 with LiDAR to perform autonomous indoor navigation and mapping.
Code - mini_pupper_navigation package
All of these developments have been pushed upstream to the MangDang Robotics Club ROS repository and are available for anyone to explore with their own Mini Pupper 2.
I’d be happy to discuss implementation details or hear from others experimenting with ROS 2 on small quadruped platforms.
Best regards,
Kishan Grewal