Generating Actionable Robot Knowledge Bases by Combining 3D Scene Graphs with Robot Ontologies

Giang Nguyen; Michael Beetz; Mihai Pomarlan; Minh Nhat Vu; Nils Leusmann; Sascha Jongebloed

REVIEW 4 major objections 4 minor 46 references

Generating Actionable Robot Knowledge Bases by Combining 3D Scene Graphs with Robot Ontologies

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Scene graphs become robot knowledge graphs via USD

desk verdict A credible systems paper that delivers a working USD-based integration pipeline and a real-robot demo, but the load-bearing semantic labeling step is unmeasured, so the 'actionable knowledge' claim needs referee pressure before acceptance. read the letter →

arxiv 2507.11770 v1 pith:BYMRFPCK submitted 2025-07-15 cs.RO

Giang Nguyen , Mihai Pomarlan , Sascha Jongebloed , Nils Leusmann , Minh Nhat Vu , Michael Beetz This is my paper

classification cs.RO

keywords UniversalSceneDescriptiongraphsrobotontologiesknowledgesemanticlabelingreportingcompetencyquestionshouseholdrobotics

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper aims to show that the many incompatible formats describing robot environments—URDF for robots, MJCF for objects, SDF and procedural JSON for rooms—can be merged into one Universal Scene Description (USD) scene graph, automatically annotated with ontology concepts, and translated into a knowledge graph a robot can query. The authors test this by converting a ProcTHOR apartment, a URDF robot, and MJCF objects into a single USD scene, running semantic reporting to link object names to knowledge repositories, and answering five competency questions that drive a real table-setting demonstration. The payoff, if the approach holds, is that raw environment data becomes actionable knowledge without manual semantic labeling, enabling robots to resolve underdetermined commands like "prepare breakfast."

What carries the argument

The load-bearing mechanism is the USD-based scene graph used as a neutral interchange format, paired with the semantic reporting step that attaches ontology classes to scene nodes. The formalization represents any world as bodies with child bodies, joints, geometries, and properties; typed schemas and API schemas in USD preserve format-specific data such as original mesh paths and dynamic parameters, allowing reverse conversion. Semantic reporting preprocesses a prim's name and type, sends it to a text-to-triples parser with the context "an X in a room," uses DBPedia links when found, and otherwise falls back to string matching against SOMA DFL concepts; the resulting semantic reports are stored in a USD layer and visualized in a web-based labeling tool before the scene is translated into the knowledge graph.

What would settle it

Rename every USD prim in the test scene to opaque identifiers such as "prim_001" while preserving geometry, materials, and layout, then rerun the pipeline; if the semantic reporting step produces no instanceOf links and the five competency questions no longer yield the answers used in the table-setting demo, the automatic-labeling claim stands falsified.

Watch

Extended reading notes

Core claim

In the paper's own terms, the central claim is that a formalized scene graph model—world, bodies, joints, geometries, and typed properties—can act as a translation layer: any common robot scene description can be stored in USD without losing critical attributes, and the USD scene can then be semantically enriched with instanceOf connections to classes in robot ontologies and knowledge graphs. The pipeline's semantic reporting step turns USD prim names into semantic reports by asking a text-to-triples tool to parse "an X in a room" and hop from DBPedia to other repositories, with a fallback that matches names against household concepts; these reports become semantic labels, and the annotated scene is converted into a KnowRob knowledge graph. The evaluation answers five competency questions about breakfast items, storage locations, tool locations, grasp surfaces, and table placement, and uses the answers to execute a real table-setting task.

Load-bearing premise

The automatic semantic labeling assumes that object names in a scene are meaningful English-like strings, so that a text parser can find a knowledge-base link or a fallback match; if objects carry arbitrary identifiers, the knowledge graph cannot be built.

Editorial extensions

If this is right

A single USD scene can combine a robot, a room, and objects originally described in different formats, and can be exported back to those formats without losing critical attributes.
Semantic reports can replace much of manual semantic labeling by deriving instanceOf links from object names alone, cutting the effort needed to build a robot knowledge base.
Competency questions for household tasks—what is needed for breakfast, where tools are stored, where to grasp an object—can be answered from the generated knowledge graph and used directly in plan execution.
Because the reasoner is rule-based, the same input always produces the same answers, giving a reproducibility that probabilistic text models do not offer for symbolic task planning.
The pipeline generalizes to any scene description format that can be expressed in the formalized scene graph model, not just the three formats tested.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper, the same USD interchange layer could ingest real-world scene graphs produced by robot perception systems, provided object names are normalized to everyday nouns before semantic reporting.
The paper leaves implicit that the fallback heuristics' success rate is measurable: running semantic reporting on scenes with deliberately uninformative names would quantify how often automatic labeling fails and where human correction is needed.
The authors' rule-based-versus-LLM comparison suggests a testable extension: benchmark plan success and consistency of the knowledge-graph reasoner against an LLM-based planner on the same table-setting task and across multiple scene layouts.
The MJCF-to-USD-to-MJCF round trip indicates USD could serve as a long-term archival format for robot scenes, a use case the paper mentions only through the transitivity test.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a solid systems paper that does what it claims, but the central semantic labeling step is not quantitatively evaluated, so the strength of the claim about 'actionable knowledge' is weaker than the demo suggests.

The contribution is real, if incremental. The paper takes scene descriptions from MJCF, URDF, and ProcTHOR, standardizes them into USD, links node names to ontology classes via a semantic-reporting pipeline (FRED plus a SOMA DFL heuristic), and feeds the resulting knowledge graph into KnowRob to answer competency questions that drive a real table-setting task. The formalization in Section IV is careful, the Docker/Binder reproducibility is a concrete plus, and the authors are honest about the key assumption: prim names need to be informative strings for humans. That assumption is stated explicitly in Section V, which is more than many systems papers do.

What is genuinely new is the specific combination: the prior ICRA 2024 paper formalized USD-to-KG translation, and this one adds the FRED-based semantic reporting, the web annotation tool, and the real-robot validation. None of these are individually novel, but the integration is useful for the semantic mapping community.

The soft spots are real but not fatal. The biggest is that semantic reporting accuracy is never measured. There is no coverage rate, no precision@k, no failure analysis for the FRED/heuristic step, and no study of how it degrades when object names are arbitrary identifiers. The web tool allows manual correction, which is honest but also means the pipeline is not as automatic as the title implies. The transitivity test (MJCF->USD->MJCF) reports similarity without a numeric measure, and the LLM comparison in Section VI-B is an assertion, not an experiment. Those are the three places a referee should push.

On the circularity point: yes, the competency questions are answered using SOMA DFL, an ontology from the authors' ecosystem, but the queries are reasonable and the real-robot demo provides some external anchor. I would not block on that.

Bottom line: the paper is for robotics researchers working on semantic maps, scene graphs, and knowledge-based reasoning. It deserves a serious referee, but the referee should ask for quantitative evaluation of the semantic labeling step and a clearer separation of automatic vs. human-assisted operation. I would engage with it.

Referee Report

4 major / 4 minor

Summary. The paper proposes a pipeline that converts heterogeneous robot scene descriptions (ProcTHOR JSON, URDF, MJCF) into a unified USD scene graph, semantically annotates USD prims using a semantic-reporting process based on FRED text-to-triples and a SOMA DFL fallback heuristic, and translates the annotated graph into a KnowRob knowledge graph. The authors claim this enables answering five competency questions that support a real table-setting robot demo. The evaluation consists of a parser experiment that unifies a ProcTHOR apartment, a URDF robot, and MJCF objects into one USD scene, an MJCF-to-USD-to-MJCF transitivity test, and a symbolic task-planning demo with a real robot. The paper also presents a web-based visualization tool for manual semantic-label correction. The central claim is that this pipeline can automatically turn scene graphs into actionable robot knowledge; the weak point is that the semantic-reporting step is not quantitatively evaluated.

Significance. If the pipeline performs reliably, it addresses a real and current integration bottleneck in robotics: scene descriptions in different formats cannot be uniformly queried or reasoned over. The use of USD as an interchange format, the reuse of the KnowRob/SOMA ecosystem, and the provision of a Docker/BinderHub setup for reproducibility are concrete strengths. However, the evidence for the central claim is mostly qualitative. The semantic-reporting step is load-bearing and is not measured for coverage or correctness; the transitivity test lacks a quantitative similarity measure; and the LLM-comparison claim is asserted without an experiment. These gaps currently prevent the paper from substantiating its conclusion that the pipeline 'proves valuable for querying in household tasks.'

major comments (4)

[Section V] The load-bearing step of the pipeline is the automatic generation of instanceOf links from USD prim names via FRED using the prompt 'an X in a room' and the SOMA DFL fallback heuristic. No coverage rate, precision, or comparison against a ground-truth object-class labeling is reported for the ProcTHOR scene or the table-setting scene. Since CQ1–CQ5 depend on these links, the conclusion that semantic reporting 'simplifies the semantic labeling process' is unsupported. Moreover, the web-based tool allows manual correction, so the amount of human effort required is unknown; if the automatic step has low accuracy on typical scene names, the pipeline's actionable-knowledge claim fails.
[Section VI-A] The transitivity test states that an open-source MJCF model is converted to USD and then back to MJCF to 'verify the models' similarity,' but no quantitative similarity measure is provided, no list of checked attributes (joint types, joint limits, inertias, mesh paths, material properties) is given, and no failure analysis is presented. Without these, the claim that the parser 'effectively maintains information integrity across different formats' is not established.
[Section VI-B] The final paragraph of Section VI-B asserts that the rule-based reasoner is 'more structured and interpretable than using a Large Language Model' and that knowledge graphs 'minimize hallucinations,' but no LLM baseline experiment, metric, or comparison is presented. This is a comparative claim about reliability and interpretability and should either be removed or supported by measurements.
[Section VI-B] The competency questions CQ1–CQ5 are formulated using predicates such as useMatch, hasDisposition, and hasPartType that come from SOMA DFL, an ontology authored within the same research group. The evaluation therefore demonstrates that the pipeline can answer hand-crafted queries against the authors' own ontology, but it does not yet demonstrate that the generated knowledge is actionable for independently specified tasks. The paper should acknowledge this closed-loop dependency and ideally include at least one competency question or task derived from an external benchmark or independent task specification.

minor comments (4)

[Section IV-B] The introduction of 'advanced typed schemas and API schemas' is vague; a small worked example showing how a URDF or MJCF property is encoded into USD would help readers reproduce or assess the mapping.
[Section VI-A] The reported loading-time improvement ('more than 10 seconds to less than 1 second') needs a specification of the scene and hardware, and it should be clarified whether this measurement is part of the transitivity evaluation or a separate optimization result.
[References] Reference [15] contains typos in its title ('cognition-enbabled robtics'); the title and spelling should be corrected.
[Listing 1] The pseudocode in Listing 1 uses 'or (...)' line breaks with inconsistent parentheses; please make the syntax uniform so the query structure is unambiguous.

Circularity Check

1 steps flagged · score 3.0 of 10

Fallback semantic labeling and competency questions share the SOMA DFL vocabulary, making part of the CQ validation a read-back of labels the pipeline itself assigned; the USD translation and FRED/DBPedia path remain independent.

fitted input called prediction [Section V (Labeling Scene Graphs with Semantic Reports), paragraph 3, and Section VI-B (Knowledge Graph for Symbolic Task Planning), CQ1/Listing 1.]
"Semantic reporting attempts to automatically identify instanceOf connections between objects in a scene and classes defined in knowledge graphs such as WikiData or ConceptNet, and ontologies such as SOMA DFL for objects and their typical uses. ... A USD prim's name and type (if given) are used to obtain the instanceOf links. ... If FRED cannot find a DBPedia link, we use a heuristic that matches the object name against a list of words associated with the concepts in SOMA DFL. ... CQ1 Which objects do I need for breakfast? ..."

For any prim whose name fails the FRED/DBPedia route, the pipeline generates its instanceOf links by matching the prim name against word lists associated with SOMA DFL concepts. The competency questions then query exactly the SOMA DFL class vocabulary (e.g., 'breakfast_food') and DFL-specific relations (useMatch, hasDisposition). Consequently, for fallback-labeled objects, a successful CQ answer is not an independent confirmation of the semantic labeling; it is a retrieval of labels assigned by the same SOMA DFL word-list matching mechanism. The FRED/DBPedia and external-knowledge branches provide independent grounding, so the circularity is partial and does not affect the USD translation or format-conversion claims.

full rationale

The core scene-graph-to-USD translation and the reverse-conversion tests are self-contained and not circular: they compare formats and preserve attributes through independent parser implementations. The FRED-based semantic reporting path also draws on external repositories (DBPedia, ConceptNet, CSKG), which gives independent content to the labeling claim. The partial circularity arises only where the fallback heuristic labels objects from SOMA DFL word lists and the evaluation queries are then written in SOMA DFL vocabulary, so part of the demonstrated query-answering success is a read-back of the system's own label assignments. The web tool's manual correction option weakens the 'automatic' claim but is a limitation rather than circularity. Overall, the central pipeline claim retains independent content, and the circularity score is moderate.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The pipeline introduces no free numeric parameters or new physical entities. Its load-bearing assumptions are about data quality (complete environment data, unique and informative prim names) and about the reliability of external text-to-triple linking and the preloaded SOMA DFL ontology. These assumptions, rather than fitted constants, are what the evaluation would need to test.

assumptions (5)

domain assumption The robot has access to all data of the environment and each USD prim has a unique name.
Stated in Section IV-B: "this paper concentrates on the semantic reasoning of the scene graphs, operating under the assumption that the robot has access to all data of the environment and that each prim ... has a unique name." This assumption excludes non-rigid-body and lighting data from the pipeline.
domain assumption Prim names and types are informative human-readable strings from which semantics can be recovered.
Section V states: "We assume that a prim name, and even more so its type, will be strings that are informative for humans." The whole semantic reporting chain depends on this.
domain assumption FRED, a text-to-triples tool, can parse "an X in a room" and return a DBPedia link for typical object names.
Section V declares the semantic reporting pipeline relies on FRED to obtain DBPedia links; if FRED fails, a heuristic match against SOMA DFL is used.
domain assumption The SOMA DFL ontology contains concepts for the household objects and use matches needed to answer the competency questions.
Section VI-B evaluates CQs that depend on predicates such as hasDisposition and useMatch, which are defined in SOMA DFL, an ontology developed by the same group.
standard math Standard math and physics tools (Mirtich mass properties, parallel axis theorem) apply to the mesh geometries.
Section IV-C invokes B. Mirtich's method and the parallel axis theorem for mass property estimation; these are established results.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Generating Actionable Robot Knowledge Bases by Combining 3D Scene Graphs with Robot Ontologies." pith.science (2026). https://pith.science/paper/BYMRFPCK

@misc{pith2026250711770,
  author       = {Pith},
  title        = {Pith review of: Generating Actionable Robot Knowledge Bases by Combining 3D Scene Graphs with Robot Ontologies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BYMRFPCK}},
  note         = {Machine review of arXiv:2507.11770}
}

read the original abstract

In robotics, the effective integration of environmental data into actionable knowledge remains a significant challenge due to the variety and incompatibility of data formats commonly used in scene descriptions, such as MJCF, URDF, and SDF. This paper presents a novel approach that addresses these challenges by developing a unified scene graph model that standardizes these varied formats into the Universal Scene Description (USD) format. This standardization facilitates the integration of these scene graphs with robot ontologies through semantic reporting, enabling the translation of complex environmental data into actionable knowledge essential for cognitive robotic control. We evaluated our approach by converting procedural 3D environments into USD format, which is then annotated semantically and translated into a knowledge graph to effectively answer competency questions, demonstrating its utility for real-time robotic decision-making. Additionally, we developed a web-based visualization tool to support the semantic mapping process, providing users with an intuitive interface to manage the 3D environment.

Figures

Figures reproduced from arXiv: 2507.11770 by the authors.

**Figure 3.** The pipeline’s architecture starts with an ambiguous command. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

**Figure 2.** Diagram of the methodology B. Deployment of the Pipeline The pipeline architecture described in this paper is built upon existing workflows. It starts with an ambiguous command such as ”Please prepare a breakfast for me,” which a natural language processor converts into executable tasks. These tasks are executed by the plan execution, which loads environmental data, including the robot and its surroundings, to crea… view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Storing various scene descriptions into USD and converting them [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: A snapshot of the web-based visualization tool displaying a USD [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

**Figure 6.** Figure 6: Overview of the semantic reporting pipeline. The text-to-triples tool [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

**Figure 7.** Figure 7: Knowledge Graph for Symbolic Task Planning in a Real-World [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 42 canonical work pages

[1]

Wordnet: An electronic lexical database , Language, Speech, and Communication, MIT Press, Cambridge, MA, 1998

work page 1998
[2]

10, 791–795

AR Abdulghany, Generalization of parallel axis theorem for rotational inertia, American Journal of Physics 85 (2017), no. 10, 791–795

work page 2017
[3]

Web 8 (2017), no

Harith Alani, Aldo Gangemi, Valentina Presutti, Diego Reforgiato Re- cupero, Andrea Giovanni Nuzzolese, Francesco Draicchio, and Misael Mongiov`ı, Semantic web machine reading with fred , Semant. Web 8 (2017), no. 6, 873–893

work page 2017
[4]

Mauricio Almeida, Jonathan Bona, Mathias Brochhausen, Werner Ceusters, Melanie Courtot, Randall Dipert, Albert Goldfain, Pierre Grenon, Janna Hastings, William Hogan, et al., Bfo 2.0 specification and user guide , 2015

work page 2015
[5]

Michael Beetz, Daniel Beßler, Andrei Haidu, Mihai Pomarlan, Asil Kaan Bozcuo ˘glu, and Georg Bartels, Knowrob 2.0 — a 2nd generation knowledge processing framework for cognition-enabled robotic agents, 2018 IEEE International Conference on Robotics and Automation (ICRA), 2018, pp. 512–519

work page 2018
[6]

Michael Beetz, Gayane Kazhoyan, and David Vernon, The CRAM cognitive architecture for robot manipulation in everyday activities , CoRR abs/2304.14119 (2023)

arXiv 2023
[7]

1012–1017

Michael Beetz, Lorenz M ¨osenlechner, and Moritz Tenorth, CRAM - A Cognitive Robot Abstract Machine for Everyday Manipulation in Human Environments, Proceedings of the 2nd IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2010) (Taipei, Taiwan), IEEE, 2010, pp. 1012–1017

work page 2010
[8]

41–64, Springer International Publishing, Cham, 2022

Michael Beetz, Simon Stelter, Daniel Beßler, Kaviya Dhanabalachan- dran, Michael Neumann, Patrick Mania, and Andrei Haidu, Robots collecting data: Modelling stores , pp. 41–64, Springer International Publishing, Cham, 2022

work page 2022

Show all 46 references

[9]

Michael Beetz, Moritz Tenorth, Dejan Pangercic, and Benjamin Pitzer, Semantic object maps for household tasks , 5th International Confer- ence on Cognitive Systems (CogSys 2012), 2012

2012
[10]

Daniel Beßler, Robert Porzel, Mihai Pomarlan, Abhijit Vyas, Sebas- tian H ¨offner, Michael Beetz, Rainer Malaka, and John A. Bateman, Foundations of the socio-physical model of activities (SOMA) for autonomous robotic agents, Formal Ontology in Information Systems - Proceeding...

2021
[11]

Knowledge Representation and Reasoning in Robotics, AAAI Spring Symposium, Citeseer, 2014

Roberto Capobianco, Guglielmo Gemignani, Daniele Nardi, Domenico Daniele Bloisi, Luca Iocchi, et al., Knowledge-based reasoning on semantic maps , Proc. Knowledge Representation and Reasoning in Robotics, AAAI Spring Symposium, Citeseer, 2014

2014
[12]

Zoey Chen, Aaron Walsman, Marius Memmel, Kaichun Mo, Alex Fang, Karthikeya Vemuri, Alan Wu, Dieter Fox, and Abhishek Gupta, Urdformer: A pipeline for constructing articulated simulation envi- ronments from real-world images , arXiv preprint arXiv:2405.11656 (2024)

2024 arXiv
[13]

of the 9th Int

Optimising Ontology Classification, Birte Glimm and Ian Horrocks and Boris Motik and Giorgos Stoilos , Proc. of the 9th Int. Seman- tic Web Conf. (ISWC 2010) (Shanghai, China), LNCS, vol. 6496, Springer, November 7–11 2010, pp. 225–240

2010
[14]

js framework, Beginning WebGL for HTML5 (2012), 173–203

Brian Danchilla and Brian Danchilla, Three. js framework, Beginning WebGL for HTML5 (2012), 173–203

2012
[15]

Jonas Dech, Pycram: A python framework for cognition-enbabled robtics
[16]

Matt Deitke, Eli VanderBilt, Alvaro Herrasti, Luca Weihs, Jordi Sal- vador, Kiana Ehsani, Winson Han, Eric Kolve, Ali Farhadi, Aniruddha Kembhavi, and Roozbeh Mottaghi, Procthor: Large-scale embodied ai using procedural generation, 2022

2022
[17]

Yan Ding, Xiaohan Zhang, Saeid Amiri, Nieqing Cao, Hao Yang, Chad Esselink, and Shiqi Zhang, Robot task planning and situation handling in open worlds , arXiv preprint arXiv:2210.01287 (2022)

2022 arXiv
[18]

George Elkoura, Sebastian Grassia, Sunya Boonyatera, Alex Mohr, Pol Jeremias-Vila, and Matt Kuruc, A deep dive into universal scene description and hydra, ACM SIGGRAPH 2019 Courses, ACM, 2019, pp. 1–48

2019
[19]

11, 955–966

Cipriano Galindo, Juan-Antonio Fern ´andez-Madrigal, Javier Gonz´alez, and Alessandro Saffiotti, Robot task planning using semantic maps , Robotics and autonomous systems 56 (2008), no. 11, 955–966

2008
[20]

Filip Ilievski, Pedro Szekely, and Bin Zhang, Cskg: The commonsense knowledge graph, Extended Semantic Web Conference (ESWC), 2021

2021
[21]

Mikhail Ivanou, Stanislav Mikhel, and Sergei Savin, Robot description formats and approaches, 2021 International Conference” Nonlinearity, Information and Robotics”(NIR), IEEE, 2021, pp. 1–5

2021
[22]

Tatiana D Kainova, Overview of the accelerated platform for robotics and artificial intelligence nvidia isaac , 2023 Seminar on Information Computing and Processing (ICP), IEEE, 2023, pp. 89–93

2023
[23]

Hyeong Ryeol Kam, Sung-Ho Lee, Taejung Park, and Chang-Hun Kim, Rviz: a toolkit for real domain data visualization , Telecommu- nication Systems 60 (2015), 337–345

2015
[24]

13699, Springer, 2022, pp

Yash Kant, Arun Ramachandran, Sriram Yenamandra, Igor Gilitschen- ski, Dhruv Batra, Andrew Szot, and Harsh Agrawal, Housekeep: Tidying virtual households using commonsense reasoning , Computer Vision - ECCV 2022 - 17th European Conference, Tel Aviv, Israel, October 23-27, 2022...

2022
[25]

Eric Kolve, Roozbeh Mottaghi, Winson Han, Eli VanderBilt, Luca Weihs, Alvaro Herrasti, Matt Deitke, Kiana Ehsani, Daniel Gordon, Yuke Zhu, et al., Ai2-thor: An interactive 3d environment for visual ai, arXiv preprint arXiv:1712.05474 (2017)

2017 arXiv
[26]

Michaela K ¨umpel, Actionable knowledge graphs - how daily activity applications can benefit from embodied web knowledge , 2024

2024
[27]

Michaela K ¨umpel, Christian A. Mueller, and Michael Beetz, Semantic digital twins for retail logistics , Dynamics in Logistics: Twenty-Five Years of Interdisciplinary Logistics Research in Bremen, Germany (Michael Freitag, Herbert Kotzab, and Nicole Megow, eds.), Springer Int...

2021
[28]

Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick van Kleef, S ¨oren Auer, and Christian Bizer, Dbpedia - a large-scale, multilingual knowledge base extracted from wikipedia

Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kon- tokostas, Pablo N. Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick van Kleef, S ¨oren Auer, and Christian Bizer, Dbpedia - a large-scale, multilingual knowledge base extracted from wikipedia. , Semantic We...

2015
[29]

2, 31–50

Brian Mirtich, Fast and accurate computation of polyhedral mass properties, Journal of graphics tools 1 (1996), no. 2, 31–50

1996
[30]

9389–9395

Giang Hoang Nguyen, Daniel Beßler, Simon Stelter, Mihai Pomar- lan, and Michael Beetz, Translating universal scene descriptions into knowledge graphs for robotic environment , 2024 IEEE Interna- tional Conference on Robotics and Automation (ICRA), IEEE, 2024, pp. 9389–9395

2024
[31]

Benjamin Nicoll, Brendan Keogh, Benjamin Nicoll, and Brendan Keogh, The unity game engine and the circuits of cultural software , Springer, 2019

2019
[32]

Arthur Nied ´zwiecki, Sascha Jongebloed, Yanxiang Zhan, Michaela K¨umpel, J ¨orn Syrbe, and Michael Beetz, Cloud-based digital twin for cognitive robotics , 2024

2024
[33]

Adam Pease, Ontology: A practical guide , Articulate Software Press, Angwin, CA, 2011

2011
[34]

Mihai Pomarlan and Robert Porzel, Narrative objects , Proceedings of the IJCAI/ECAI Workshop on Semantic Techniques for Narrative- based Understanding, 07 2022

2022
[35]

Edson Prestes, Joel Luis Carbonera, Sandro Rama Fiorini, Vitor A. M. Jorge, Mara Abel, Raj Madhavan, Angela Locoro, Paulo Goncalves, Marcos E. Barreto, Maki Habib, Abdelghani Chibani, S ´ebastien G´erard, Yacine Amirat, and Craig Schlenoff, Towards a core ontology for robotics...

2013
[36]

Xavier Puig, Eric Undersander, Andrew Szot, Mikael Dallaire Cote, Tsung-Yen Yang, Ruslan Partsey, Ruta Desai, Alexander William Clegg, Michal Hlavac, So Yeon Min, et al., Habitat 3.0: A co-habitat for humans, avatars and robots , arXiv preprint arXiv:2310.13724 (2023)

2023 arXiv
[37]

4444–4451

Robyn Speer, Joshua Chin, and Catherine Havasi, Conceptnet 5.5: an open multilingual graph of general knowledge , Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence, AAAI’17, AAAI Press, 2017, p. 4444–4451

2017
[38]

Andreas Steigmiller, Thorsten Liebig, and Birte Glimm, Konclude: System description, Journal of Web Semantics (JWS) 27 (2014), 78– 85

2014
[39]

Niko S ¨underhauf, Trung T. Pham, Yasir Latif, Michael Milford, and Ian Reid, Meaningful maps with object-oriented semantic mapping , 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017, pp. 5079–5085

2017
[40]

5, 566–590

Moritz Tenorth and Michael Beetz, Knowrob: A knowledge processing infrastructure for cognition-enabled robots , The International Journal of Robotics Research 32 (2013), no. 5, 566–590

2013
[41]

Moritz Tenorth, Lars Kunze, Dominik Jain, and Michael Beetz, Knowrob-map - knowledge-linked semantic object maps , 2010 10th IEEE-RAS International Conference on Humanoid Robots, 2010, pp. 430–435

2010
[42]

5026–5033

Emanuel Todorov, Tom Erez, and Yuval Tassa, Mujoco: A physics en- gine for model-based control, 2012 IEEE/RSJ international conference on intelligent robots and systems, IEEE, 2012, pp. 5026–5033

2012
[43]

ACM 57 (2014), no

Denny Vrande ˇci´c and Markus Kr¨otzsch, Wikidata: A free collaborative knowledgebase, Commun. ACM 57 (2014), no. 10, 78–85

2014
[44]

Xinyue Wei, Minghua Liu, Zhan Ling, and Hao Su, Approximate convex decomposition for 3d meshes with collision-aware concavity and tree search , ACM Transactions on Graphics (TOG) 41 (2022), no. 4, 1–18

2022
[45]

5922–5931

Luca Weihs, Matt Deitke, Aniruddha Kembhavi, and Roozbeh Mot- taghi, Visual room rearrangement , Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2021, pp. 5922–5931

2021
[46]

Maria Keet, Analysis of ontology competency questions and their formalizations in sparql-owl , Journal of Web Semantics 59 (2019), 100534

Dawid Wi ´sniewski, Jedrzej Potoniec, Agnieszka Ławrynowicz, and C. Maria Keet, Analysis of ontology competency questions and their formalizations in sparql-owl , Journal of Web Semantics 59 (2019), 100534

2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Wordnet: An electronic lexical database , Language, Speech, and Communication, MIT Press, Cambridge, MA, 1998

work page 1998

[2] [2]

10, 791–795

AR Abdulghany, Generalization of parallel axis theorem for rotational inertia, American Journal of Physics 85 (2017), no. 10, 791–795

work page 2017

[3] [3]

Web 8 (2017), no

Harith Alani, Aldo Gangemi, Valentina Presutti, Diego Reforgiato Re- cupero, Andrea Giovanni Nuzzolese, Francesco Draicchio, and Misael Mongiov`ı, Semantic web machine reading with fred , Semant. Web 8 (2017), no. 6, 873–893

work page 2017

[4] [4]

Mauricio Almeida, Jonathan Bona, Mathias Brochhausen, Werner Ceusters, Melanie Courtot, Randall Dipert, Albert Goldfain, Pierre Grenon, Janna Hastings, William Hogan, et al., Bfo 2.0 specification and user guide , 2015

work page 2015

[5] [5]

Michael Beetz, Daniel Beßler, Andrei Haidu, Mihai Pomarlan, Asil Kaan Bozcuo ˘glu, and Georg Bartels, Knowrob 2.0 — a 2nd generation knowledge processing framework for cognition-enabled robotic agents, 2018 IEEE International Conference on Robotics and Automation (ICRA), 2018, pp. 512–519

work page 2018

[6] [6]

Michael Beetz, Gayane Kazhoyan, and David Vernon, The CRAM cognitive architecture for robot manipulation in everyday activities , CoRR abs/2304.14119 (2023)

arXiv 2023

[7] [7]

1012–1017

Michael Beetz, Lorenz M ¨osenlechner, and Moritz Tenorth, CRAM - A Cognitive Robot Abstract Machine for Everyday Manipulation in Human Environments, Proceedings of the 2nd IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2010) (Taipei, Taiwan), IEEE, 2010, pp. 1012–1017

work page 2010

[8] [8]

41–64, Springer International Publishing, Cham, 2022

Michael Beetz, Simon Stelter, Daniel Beßler, Kaviya Dhanabalachan- dran, Michael Neumann, Patrick Mania, and Andrei Haidu, Robots collecting data: Modelling stores , pp. 41–64, Springer International Publishing, Cham, 2022

work page 2022