Using Abstraction for Interpretable Robot Programs in Stochastic Domains. Hofmann, T. & Belle, V. In KR Workshop on Explainable Logic-Based Knowledge Representation (XLoKR), 2022. doi abstract bibtex 38 downloads A robot's actions are inherently stochastic, as its sensors are noisy and its actions do not always have the intended effects. For this reason, the agent language Golog has been extended to models with degrees of belief and stochastic actions. While this allows more precise robot models, the resulting programs are much harder to comprehend, because they need to deal with the noise, e.g., by looping until some desired state has been reached with certainty, and because the resulting action traces consist of a large number of actions cluttered with sensor noise. To alleviate these issues, we propose to use abstraction. We define a high-level and nonstochastic model of the robot and then map the high-level model into the lower-level stochastic model. The resulting programs are much easier to understand, often do not require belief operators or loops, and produce much shorter action traces.
@inproceedings{hofmannUsingAbstractionInterpretable2022,
entrysubtype = {workshop},
title = {Using Abstraction for Interpretable Robot Programs in Stochastic Domains},
booktitle = {{{KR Workshop}} on {{Explainable Logic-Based Knowledge Representation}} ({{XLoKR}})},
author = {Hofmann, Till and Belle, Vaishak},
year = 2022,
doi = {10.48550/arXiv.2207.12763},
abstract = {A robot's actions are inherently stochastic, as its sensors are noisy and its actions do not always have the intended effects. For this reason, the agent language Golog has been extended to models with degrees of belief and stochastic actions. While this allows more precise robot models, the resulting programs are much harder to comprehend, because they need to deal with the noise, e.g., by looping until some desired state has been reached with certainty, and because the resulting action traces consist of a large number of actions cluttered with sensor noise. To alleviate these issues, we propose to use abstraction. We define a high-level and nonstochastic model of the robot and then map the high-level model into the lower-level stochastic model. The resulting programs are much easier to understand, often do not require belief operators or loops, and produce much shorter action traces.},
contribution = {This paper demonstrates how abstraction can be used to increase the interpretability of robot programs and is directly tied to \cite{hofmannAbstractingNoisyRobot2023}. The general connection between abstraction and explainability is due to Vaishak Belle. The idea to apply the concepts from \cite{hofmannAbstractingNoisyRobot2023} to increase interpretability was by me. I wrote major parts of the paper.},
copyright = {All rights reserved},
credit = {{Methodology, Formal analysis, Investigation, Writing - Original Draft, Writing - Review \& Editing, Visualization}},
publicationtype = {{original research article}},
shorthand = {XLoKR22}
}
Downloads: 38
{"_id":"rGeLXdDhnjgDgbk7z","bibbaseid":"hofmann-belle-usingabstractionforinterpretablerobotprogramsinstochasticdomains-2022","author_short":["Hofmann, T.","Belle, V."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","entrysubtype":"workshop","title":"Using Abstraction for Interpretable Robot Programs in Stochastic Domains","booktitle":"KR Workshop on Explainable Logic-Based Knowledge Representation (XLoKR)","author":[{"propositions":[],"lastnames":["Hofmann"],"firstnames":["Till"],"suffixes":[]},{"propositions":[],"lastnames":["Belle"],"firstnames":["Vaishak"],"suffixes":[]}],"year":"2022","doi":"10.48550/arXiv.2207.12763","abstract":"A robot's actions are inherently stochastic, as its sensors are noisy and its actions do not always have the intended effects. For this reason, the agent language Golog has been extended to models with degrees of belief and stochastic actions. While this allows more precise robot models, the resulting programs are much harder to comprehend, because they need to deal with the noise, e.g., by looping until some desired state has been reached with certainty, and because the resulting action traces consist of a large number of actions cluttered with sensor noise. To alleviate these issues, we propose to use abstraction. We define a high-level and nonstochastic model of the robot and then map the high-level model into the lower-level stochastic model. The resulting programs are much easier to understand, often do not require belief operators or loops, and produce much shorter action traces.","contribution":"This paper demonstrates how abstraction can be used to increase the interpretability of robot programs and is directly tied to i̧tehofmannAbstractingNoisyRobot2023. The general connection between abstraction and explainability is due to Vaishak Belle. The idea to apply the concepts from i̧tehofmannAbstractingNoisyRobot2023 to increase interpretability was by me. I wrote major parts of the paper.","copyright":"All rights reserved","credit":"Methodology, Formal analysis, Investigation, Writing - Original Draft, Writing - Review & Editing, Visualization","publicationtype":"original research article","shorthand":"XLoKR22","bibtex":"@inproceedings{hofmannUsingAbstractionInterpretable2022,\n entrysubtype = {workshop},\n title = {Using Abstraction for Interpretable Robot Programs in Stochastic Domains},\n booktitle = {{{KR Workshop}} on {{Explainable Logic-Based Knowledge Representation}} ({{XLoKR}})},\n author = {Hofmann, Till and Belle, Vaishak},\n year = 2022,\n doi = {10.48550/arXiv.2207.12763},\n abstract = {A robot's actions are inherently stochastic, as its sensors are noisy and its actions do not always have the intended effects. For this reason, the agent language Golog has been extended to models with degrees of belief and stochastic actions. While this allows more precise robot models, the resulting programs are much harder to comprehend, because they need to deal with the noise, e.g., by looping until some desired state has been reached with certainty, and because the resulting action traces consist of a large number of actions cluttered with sensor noise. To alleviate these issues, we propose to use abstraction. We define a high-level and nonstochastic model of the robot and then map the high-level model into the lower-level stochastic model. The resulting programs are much easier to understand, often do not require belief operators or loops, and produce much shorter action traces.},\n contribution = {This paper demonstrates how abstraction can be used to increase the interpretability of robot programs and is directly tied to \\cite{hofmannAbstractingNoisyRobot2023}. The general connection between abstraction and explainability is due to Vaishak Belle. The idea to apply the concepts from \\cite{hofmannAbstractingNoisyRobot2023} to increase interpretability was by me. I wrote major parts of the paper.},\n copyright = {All rights reserved},\n credit = {{Methodology, Formal analysis, Investigation, Writing - Original Draft, Writing - Review \\& Editing, Visualization}},\n publicationtype = {{original research article}},\n shorthand = {XLoKR22}\n}\n\n","author_short":["Hofmann, T.","Belle, V."],"key":"hofmannUsingAbstractionInterpretable2022","id":"hofmannUsingAbstractionInterpretable2022","bibbaseid":"hofmann-belle-usingabstractionforinterpretablerobotprogramsinstochasticdomains-2022","role":"author","urls":{},"metadata":{"authorlinks":{}},"downloads":38},"bibtype":"inproceedings","biburl":"https://ml.rwth-aachen.de/~till.hofmann/files/publications.bib","dataSources":["dqRQPSg6Hy3ZXQg7z","ayruv4JDJNMLz7ieC","o8SJynH3ZqhBLq3sp","qb6u34NNzEHCho66A","7PCTuJcihysb7xARD","P2RLsqAnkwjGHmpeQ","zp5TdwDdbpGvos9cF","r4qCqi5wRBiYBfSgt"],"keywords":[],"search_terms":["using","abstraction","interpretable","robot","programs","stochastic","domains","hofmann","belle"],"title":"Using Abstraction for Interpretable Robot Programs in Stochastic Domains","year":2022,"downloads":38}