2026-08-06T15:00:00Z
This year the Open-RMF project participated in Google Summer of Code with two contributors to help with different aspects of the Next Generation initiative. At this session of the Interoperability Special Interest Group, these contributors will present what they’ve done so far.
1. Zhen Wan: Crossflow Diagram Editor
Crossflow is a recently developed flow-based programming library aimed at making highly asynchronous service-driven systems easy to design and execute. Crossflow will be used by Open-RMF to develop extensible, modular, sophisticated, multi-agent behaviors in its Next Generation. Zhen Wan is participating in Google Summer of Code this year to introduce significant user experience improvements to Crossflow’s open source diagram editor, which allows users to graphically construct workflow diagrams which can then be executed by Crossflow. His improvements include streamlining the process of adding elements, filtering UI options based on compatibility evaluations, giving context-sensitive interaction hints to users, rendering the actions inside an active diagram, displaying the execution event trace, and improving the Python scripting feature.
Join us to witness these latest improvements to the user experience of the crossflow diagram editor.
2. Samuel Foo: Next Generation Traffic Management
Multi-robot fleets operating in shared spaces must handle both contention for limited destinations and obstacles that appear after a plan has been issued. We present two extensions to the Open-RMF next-generation prototype that address these challenges.
The first part introduces a reservation-aware destination server that selects among alternative goals, prevents overlapping assignments, and queues or parks blocked robots.
The second part presents a layered global occupancy map that combines a static map with short-lived robot observations. It preserves a standard ROS occupancy-grid interface, allowing the planning stack to detect obstacles blocking an active route and request a controlled replan.
Through demonstrations and integration tests, we show how these components address resource allocation and dynamic replanning while remaining modular and independently extensible.