Institution profile

Johannes Kepler University Linz

Academic institutioneurope · at
Official website
Research library340linked papers
Opportunities0open roles
Selected work

Representative Papers

An O(n)-Algorithm for the Higher-Order Kinematics and Inverse Dynamics of Serial Manipulators Using Spatial Representation of Twists

Apr 01, 2021IEEE Robotics and Automation Letters

To address the real-time computational demands of optimal robot control—particularly differential flatness-based control—for high-order kinematics and inverse dynamics, this paper proposes a unified algorithm grounded in spatial screw theory and Lie group/Lie algebra formalism. The method achieves linear-time complexity (O(n)) for both forward and inverse kinematics up to the fourth order, as well as second-order inverse dynamics, via recursive forward/backward propagation and rigid-body screw modeling, fully supporting vectorized parameter inputs. The algorithm is compact, analytically exact, and real-time capable. Experimental validation on a Franka Panda 7-DOF manipulator demonstrates significant speedup over conventional approaches, enabling millisecond-level response required for high-order closed-loop control. The core contribution is the first unified O(n) framework integrating fourth-order kinematics and second-order inverse dynamics, establishing a foundational advancement for real-time differential flatness control of serial manipulators.

14 citations3 influentialRead paper

Nominal Anti-Unification

Jun 29, 2015International Conference on Rewriting Techniques and Applications

This paper addresses nominal anti-unification—the computation of a least general generalization (LGG) of given terms within contexts containing binding structures. Standard first-order anti-unification fails for bound variables, and this work establishes, for the first time in the nominal syntax framework, that an LGG exists and is unique up to variable renaming and α-equivalence when the underlying atom set is finite. To compute it, we propose the first sound and complete constructive algorithm, integrating nominal logic, equivariance checking, α-equivalence handling, and context-sensitive generalization. We formally prove its polynomial-time complexity. Our approach provides a rigorous and efficient foundation for binding-aware inductive learning and code clone detection.

13 citations1 influentialRead paper

Singularity-Free Lie Group Integration and Geometrically Consistent Evaluation of Multibody System Models described in terms of Standard Absolute Coordinates

Dec 24, 2021Journal of Computational and Nonlinear Dynamics

This work proposes a Lie group integration framework compatible with the standard absolute coordinate formulation for multibody system modeling. Traditional time integration methods struggle to preserve the geometric consistency of rigid body spatial motion due to their reliance on singularity-prone absolute coordinate parameterizations in vector spaces. The proposed approach overcomes this limitation by introducing a local–global transformation (LGT) mapping that seamlessly integrates pose representations on the Lie groups SO(3)×ℝ³ or SE(3) with classical vector-space integration techniques. For the first time, this method enables Lie group integrators to be directly embedded into existing absolute-coordinate-based multibody equations without requiring any reformulation of the simulation code. Consequently, the integration remains free of singularities while preserving the underlying geometric structure, significantly enhancing both accuracy and numerical stability in simulations.

4 citations1 influentialRead paper

LiveXiv - A Multi-Modal Live Benchmark Based on Arxiv Papers Content

Oct 14, 2024arXiv.org

Test data contamination from web crawling undermines the validity of multimodal model evaluation. Method: This paper introduces the first dynamic, evolvable “living” multimodal benchmark grounded in arXiv scientific papers. It (1) achieves precise figure–text alignment via multimodal PDF/TeX parsing; (2) jointly employs rule-based heuristics and large language models (LLMs) to generate high-quality, human-annotation-free visual question answering (VQA) data; and (3) proposes an incremental benchmark evolution mechanism coupled with a statistically grounded sparse-subset evaluation algorithm for efficient full-benchmark performance estimation. Contribution/Results: Human validation confirms an automatic annotation error rate <2.5%. Comprehensive evaluation on the inaugural benchmark reveals significant, previously undetected capability gaps across leading open- and closed-source multimodal large models. The dataset is publicly available on Hugging Face; code will be released shortly.

4 citationsRead paper

Cybersecurity AI: A Game-Theoretic AI for Guiding Attack and Defense

Jan 09, 2026arXiv.org

This work addresses the strategic limitations of current AI-driven penetration testing, which lacks the intuitive decision-making capabilities of expert human adversaries in cyber conflict scenarios. To bridge this gap, the authors propose Generative Cut-the-Rope (G-CTR), a novel framework that integrates game theory with large language models (LLMs) in a closed-loop architecture. G-CTR extracts attack graphs from agent contexts, computes cost-aware Nash equilibria, and uses generative summaries as feedback to guide LLM-based reasoning toward strategic-level offensive and defensive decisions. Experimental results demonstrate that G-CTR replicates 70–90% of expert attack graph structures in real-world settings, increases attack success rates from 20.0% to 42.9%, accelerates execution by 60–245×, reduces operational costs by over 140×, decreases behavioral variance by 5.2×, and achieves Purple team win ratios ranging from 2:1 to 3.7:1.

1 citationsRead paper
Recent publications

Latest Papers