https://markusknauer.github.io/projects/Recent content in Projects on Markus KnauerHugoen-USm.knauer@tum.de (Markus Knauer)m.knauer@tum.de (Markus Knauer)Fri, 14 Aug 2026 09:50:11 +0200https://markusknauer.github.io/projects/momo/Fri, 24 Apr 2026 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/momo/MOMO unifies physical, verbal and graphical robot skill adaptation on a 7-DoF torque-controlled robot. Published in IEEE RA-P 2026.https://markusknauer.github.io/projects/clasp/Mon, 20 Apr 2026 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/clasp/CLASP combines Vision-Language Models with Task-Parameterized KMPs for language-driven robot skill selection, composition and active acquisition from 2–5 demos.https://markusknauer.github.io/projects/irosa/Sat, 14 Mar 2026 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/irosa/IROSA adapts robot skills through natural language with a tool-based LLM architecture over Kernelized Movement Primitives. Published in IEEE RA-L 2026.https://markusknauer.github.io/projects/loci/Thu, 13 Feb 2025 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/loci/LOCI: interactive incremental learning of generalizable robot skills with local trajectory modulation (TP-KMP). Published in IEEE RA-L 2025.https://markusknauer.github.io/projects/raccoon/Mon, 28 Apr 2025 08:59:00 +0200m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/raccoon/RACCOON grounds embodied question-answering in a robot’s own world and task models using RAG, raising truthful answers from 15.9% to 82.5%. IEEE ICRA 2025.https://markusknauer.github.io/projects/recall/Sun, 23 Oct 2022 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/recall/RECALL: A rehearsal-free continual learning approach for object classification achieving state-of-the-art results. Published at IEEE IROS 2022.https://markusknauer.github.io/projects/hows-cl-25/Sat, 22 Oct 2022 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/hows-cl-25/HOWS-CL-25: A synthetic dataset with 150,795 unique images across 25 household object categories for continual learning research, created using BlenderProc.https://markusknauer.github.io/projects/blenderproc/Sun, 01 Oct 2023 00:00:00 +0000m.knauer@tum.de (Markus Knauer)https://markusknauer.github.io/projects/blenderproc/BlenderProc: open source procedural pipeline for photorealistic synthetic training images — RGB, depth, segmentation, normals and poses. 3.4k+ GitHub stars.