I build the tools that let people see what a cognitive robot is doing. I work on the CRAM cognitive architecture at the Institute for Artificial Intelligence, University of Bremen — mostly on the parts where plans, knowledge and a 3D world meet in the browser.
Live demo · Repository · Changelog
- Recorded runs replay in any browser — URDFs, meshes and the real giskardpy trajectory, no ROS needed.
- Live runs stream in from a running demo: plan steps light up, motion statecharts stream, and dragging an object in the viewer moves it in the demo's world.
- Questions in English — which arm picked up the milk? — are matched to EQL queries over the recorded episode, answered, and replayed at the moment they name.
- A Plan Builder composes a new plan by drag and drop and generates a runnable demonstration.
Built with three.js, a small Python server, and the CRAM stack (coraplex, giskardpy, krrood, semantic_digital_twin).
- The CRAM monorepo: cram2/cognitive_robot_abstract_machine — cramera lives in
cramera/ - Ready-made recordings: cram2/cram-scenes