Reflections from ROSCon 2025 β Singapore πΈπ¬
Just wrapped up an amazing few days at ROSCon 2025 (October 27β29) here in Singapore β and wow, what an experience! From deep tech sessions to hallway chats with some of the brightest robotics minds, it was a reminder of how fast the ROS ecosystem is evolving.
Here are a few highlights and standout projects that truly stood out π
πΉ Key Updates
-
rmw_zenoh becomes Tier-1 Middleware
Officially elevated to Tier-1 in ROS 2 Kilted Kaiju, enabling faster, more distributed, real-time communication.
-
Open-RMF Expands in Real Deployments
Success stories at Changi General Hospital and Changi Airport showcase the power of open, coordinated multi-robot fleets.
-
New Special Interest Group (SIG) for Physical AI
A new SIG for Physical AI was announced β aiming to make ROS the foundation for data collection and deployment of visual-motor policies and physical AI models, bridging the gap between hardware, datasets, and large-scale embodied intelligence.
-
NVIDIA Isaac GROOT 1.6
Arriving mid-November, brings major upgrades β improved VLM backbone, diverse pretraining data, enhanced DiT and control, plus open-source code and weights on GitHub and Hugging Face.
Standout Projects
π§ Planning, Simulation & ML Integration
-
Robotplan β Motion planning & control toolkit for ROS 2.
-
Genesis ROS β Connects ML simulators with ROS 2 for real-time, time-accelerated learning.
-
Crossflow β Rust-based workflow engine powering Open-RMF multi-agent coordination.
-
ROS 2 MCP Server β Bridges LLMs with ROS 2 via the Model Context Protocol, enabling natural-language robot interaction.
π§© Debugging, Testing & Dev Tools
-
Replay Testing β Framework for automated tests against real robot data.
-
Rtest β Deterministic ROS 2 unit tests, eliminating flaky CI runs.
-
ros2bag_tools β Simplified command-line tools to cut, extract, and sync ROS 2 bag files.
-
Cockpit ROS 2 Diagnostics β Web-based GUI to monitor robot diagnostics and logs.
-
Custom Message Visualizer β Interactive visualization of custom ROS messages for faster debugging.
π₯ Vision & Streaming
-
Cloudini β Compresses point clouds up to 85% for efficient mapping and perception.
-
FFmpeg Pipeline for ROS 2 β Integrates multimedia streaming pipelines using FFmpeg.
-
Smarter Vision Pipelines β Tools to compress, transport, and synchronize large-scale visual data efficiently.
-
ABR Image Transport β Adaptive bitrate streaming for dynamic, network-aware video feeds.
π οΈ Workflow, Build & Management
-
Robocore CLI β Modular command-line tool for scalable ROS 2 builds and deployments.
-
launch_frontend_py β Simplifies ROS 2 launch configurations using declarative XML/YAML syntax.
-
rclrs (ROS 2 Rust) β Brings Rustβs memory safety, speed, and concurrency to ROS 2.
-
Pixi + RoboStack β Streamlined cross-platform ROS installation and dependency management.
π‘ Closing Thoughts
Every year, ROSCon reminds me why I love this field β the openness, collaboration, and shared passion for building intelligent, connected robots.
As we move toward Physical AI, standardized data collection, and cross-domain integration, ROS continues to stand out as the backbone of modern robotics research and deployment.
PS: I might have missed a few great projects β feel free to explore more highlights and talks at roscon.ros.org/2025