Cloud-Based Hierarchical Imitation Learning for Scalable Transfer of Construction Skills from Human Workers to Assisting Robots DOI Creative Commons

Hongrui Yu,

Vineet R. Kamat, Carol C. Menassa

и другие.

arXiv (Cornell University), Год журнала: 2023, Номер unknown

Опубликована: Янв. 1, 2023

Assigning repetitive and physically-demanding construction tasks to robots can alleviate human workers's exposure occupational injuries. Transferring necessary dexterous adaptive artisanal craft skills from workers is crucial for the successful delegation of achieving high-quality robot-constructed work. Predefined motion planning scripts tend generate rigid collision-prone robotic behaviors in unstructured site environments. In contrast, Imitation Learning (IL) offers a more robust flexible skill transfer scheme. However, majority IL algorithms rely on repeatedly demonstrate task performance at full scale, which be counterproductive infeasible case To address this concern, paper proposes an immersive, cloud robotics-based virtual demonstration framework that serves two primary purposes. First, it digitalizes process, eliminating need physical manipulation heavy objects. Second, employs federated collection reusable demonstrations are transferable similar future thus reduce requirement illustration by agents. Additionally, enhance trustworthiness, explainability, ethical soundness robot training, utilizes Hierarchical (HIL) model decompose into sequential reactive sub-skills. These layers represented deep generative models, enabling control actions. By delegating strains work human-trained robots, promotes inclusion with diverse capabilities educational backgrounds within industry.

Язык: Английский

Considerations on the Dynamics of Biofidelic Sensors in the Assessment of Human–Robot Impacts DOI Creative Commons
Buddhika Piyumal Bandara Samarathunga Samarathunga Mudiyanselage, Marcello Valori, Rodolfo Faglia

и другие.

Machines, Год журнала: 2023, Номер 12(1), С. 26 - 26

Опубликована: Дек. 30, 2023

Ensuring the safety of physical human–robot interaction (pHRI) is utmost importance for industries and organisations seeking to incorporate robots into their workspaces. To address this concern, ISO/TS 15066:2016 outlines hazard analysis preventive measures ensuring in Human–Robot Collaboration (HRC). analyse contact, it common practice separately evaluate “transient” “quasi-static” contact phases. Accurately measuring transient forces during close collaboration requires so-called “biofidelic” sensors that closely mimic human tissue properties, featuring adequate bandwidth balanced damping. The dynamics interactions using biofidelic devices are being explored research. In paper, one sensor tested its dynamic characteristics identify main factors influencing performance practical applications testing. aim, parameters, such as natural frequency damping coefficient, estimated by utilising a custom pendulum setup impact sensor. Mathematical models developed characterise system also disclosed.

Язык: Английский

Процитировано

2
Cloud-Based Hierarchical Imitation Learning for Scalable Transfer of Construction Skills from Human Workers to Assisting Robots DOI Creative Commons

Hongrui Yu,

Vineet R. Kamat, Carol C. Menassa

и другие.

arXiv (Cornell University), Год журнала: 2023, Номер unknown

Опубликована: Янв. 1, 2023

Assigning repetitive and physically-demanding construction tasks to robots can alleviate human workers's exposure occupational injuries. Transferring necessary dexterous adaptive artisanal craft skills from workers is crucial for the successful delegation of achieving high-quality robot-constructed work. Predefined motion planning scripts tend generate rigid collision-prone robotic behaviors in unstructured site environments. In contrast, Imitation Learning (IL) offers a more robust flexible skill transfer scheme. However, majority IL algorithms rely on repeatedly demonstrate task performance at full scale, which be counterproductive infeasible case To address this concern, paper proposes an immersive, cloud robotics-based virtual demonstration framework that serves two primary purposes. First, it digitalizes process, eliminating need physical manipulation heavy objects. Second, employs federated collection reusable demonstrations are transferable similar future thus reduce requirement illustration by agents. Additionally, enhance trustworthiness, explainability, ethical soundness robot training, utilizes Hierarchical (HIL) model decompose into sequential reactive sub-skills. These layers represented deep generative models, enabling control actions. By delegating strains work human-trained robots, promotes inclusion with diverse capabilities educational backgrounds within industry.

Язык: Английский

Процитировано

0