Between May and September 2026, eProsima presented three important developments in Vulcanexus, one of the building blocks of the ARISE framework: a modular toolkit for human-robot interaction, a more scalable way to share data about people, and a beta framework for AI-based task planning.
A robot working alongside people may need to know who is around, understand what they say and decide what to do next. Each of those capabilities depends on software that ROS 2 developers would otherwise have to build and connect themselves.
ARISE is a European project developing open technologies for human-centric robotics applications. Vulcanexus, the ROS 2 tool set developed and maintained by eProsima, is part of its technical foundation. Its latest developments address the three needs above.
Vulcanexus HRI: perception and interaction modules for ROS 2
Vulcanexus HRI is a set of modules that give a ROS 2 application information about the people around the robot, and ways to communicate with them. It covers face detection, human pose estimation, emotion recognition based on facial expressions, visual displays, speech-to-text and text-to-speech. Shared message definitions let these modules work together as one system.
Each module can run on its own or alongside the others. A team could use pose estimation to know where an operator is standing and text-to-speech to tell them a task is finished, without redesigning the rest of the application. Because Vulcanexus HRI runs within existing Vulcanexus and ROS 2 setups and builds on the latest Fast DDS features, it can be added to applications that are already running.
hri_msgs: keeping communication manageable when several people are present
Handling data about one detected person is relatively straightforward. As several people are tracked simultaneously, however, the communication architecture becomes more demanding. In many HRI designs, each detected person gets its own namespace, with its own topics, publishers and subscribers. The approach is clear, but the communication graph grows with every new person in the scene, and large ROS 2 systems can become saturated.
hri_msgs takes a different route. It relies on Fast DDS Keys, which let a single topic carry many instances of the same data type, each with its own identifier. All detected faces can share one topic and still be told apart. It also aggregates related data: a face’s region, landmarks and pose can travel in a single message instead of on separate topics.
The outcome is fewer topics and communication entities, less middleware overhead and simpler discovery. A smaller graph is easier to visualise and debug, which matters when an application moves from a demo to a real deployment.
VulcanAI: from a goal to a plan the robot can execute
Robots can already perceive, navigate, call services and execute actions. As applications grow, the harder part is deciding what to do first, what can run in parallel and how to recover when a step fails.
VulcanAI is a framework that uses language models as planning assistants. Given a request, it breaks it into sub-tasks, makes explicit which steps run in sequence and which in parallel, and selects the tools to carry them out from a registry. Steps can pass results to one another, and an iterative manager checks progress and regenerates the plan when needed.
VulcanAI comes with default ROS 2 tools for nodes, topics, services, actions, parameters and interfaces, so it can also help inspect and debug running systems. It sits on top of the existing middleware rather than replacing it. The project is in beta and under active development, so for now it is best suited to experimentation.
Why this matters for ARISE
Taken together, these developments cover different layers of the path from perceiving a person to acting on a request. For ARISE, they strengthen the middleware capabilities available for building human-centric robotic applications: Vulcanexus HRI provides reusable perception and interaction modules, hri_msgs makes human-related data exchange more manageable as applications scale, and VulcanAI extends the same ROS 2 environment with AI-based planning and tool-aware execution.
Get started
To try Vulcanexus HRI, see the Overview section of the Vulcanexus documentation
The original articles from eProsima:
- Vulcanexus HRI, building the next generation of human ware robotic systems. (28 May 2026)
- Vulcanexus HRI, a smarter way to scale human robot interaction in ros2 (18 September 2026)
- Introducing VulcanAI planning and reasoning for robotic applications (23 September 2026)
Vulcanexus continues to evolve, with new capabilities and improvements being introduced over time. We’ll be sharing regular updates as new developments become available, so stay tuned for what’s next.