Predicting Mid-Air Interaction Movements and Fatigue Using Deep Reinforcement Learning

dc.contributorAalto-yliopistofi
dc.contributorAalto Universityen
dc.contributor.authorCheema, Noshabaen_US
dc.contributor.authorFrey-Law, Laura A.en_US
dc.contributor.authorNaderi, Kouroshen_US
dc.contributor.authorLehtinen, Jaakkoen_US
dc.contributor.authorSlusallek, Philippen_US
dc.contributor.authorHämäläinen, Perttuen_US
dc.contributor.departmentMax Planck Institute for Informaticsen_US
dc.contributor.departmentUniversity of Iowaen_US
dc.contributor.departmentProfessorship Hämäläinen Perttuen_US
dc.contributor.departmentProfessorship Lehtinen Jaakkoen_US
dc.contributor.departmentSaarland Universityen_US
dc.contributor.departmentDepartment of Computer Scienceen
dc.contributor.departmentDepartment of Mediaen
dc.date.accessioned2020-12-31T08:48:50Z
dc.date.available2020-12-31T08:48:50Z
dc.date.issued2020en_US
dc.description.abstractA common problem of mid-air interaction is excessive arm fatigue, known as the "Gorilla arm" effect. To predict and prevent such problems at a low cost, we investigate user testing of mid-air interaction without real users, utilizing biomechanically simulated AI agents trained using deep Reinforcement Learning (RL). We implement this in a pointing task and four experimental conditions, demonstrating that the simulated fatigue data matches human fatigue data. We also compare two effort models: 1) instantaneous joint torques commonly used in computer animation and robotics, and 2) the recent Three Compartment Controller (3CC-) model from biomechanical literature. 3CC- yields movements that are both more efficient and relaxed, whereas with instantaneous joint torques, the RL agent can easily generate movements that are quickly tiring or only reach the targets slowly and inaccurately. Our work demonstrates that deep RL combined with the 3CC- provides a viable tool for predicting both interaction movements and user experiencein silico, without users.en
dc.description.versionPeer revieweden
dc.format.extent13
dc.format.extent1–13
dc.format.mimetypeapplication/pdfen_US
dc.identifier.citationCheema , N , Frey-Law , L A , Naderi , K , Lehtinen , J , Slusallek , P & Hämäläinen , P 2020 , Predicting Mid-Air Interaction Movements and Fatigue Using Deep Reinforcement Learning . in Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems . , 572 , ACM , New York, NY, USA , pp. 1–13 , ACM SIGCHI Annual Conference on Human Factors in Computing Systems , Honolulu , Hawaii , United States , 26/04/2020 . https://doi.org/10.1145/3313831.3376701en
dc.identifier.doi10.1145/3313831.3376701en_US
dc.identifier.isbn9781450367080
dc.identifier.otherPURE UUID: d0a15f11-0457-49ac-955b-72605ab1d8b4en_US
dc.identifier.otherPURE ITEMURL: https://research.aalto.fi/en/publications/d0a15f11-0457-49ac-955b-72605ab1d8b4en_US
dc.identifier.otherPURE LINK: https://doi.org/10.1145/3313831.3376701en_US
dc.identifier.otherPURE FILEURL: https://research.aalto.fi/files/54100001/Cheema_Predicting.3313831.3376701.pdfen_US
dc.identifier.urihttps://aaltodoc.aalto.fi/handle/123456789/101634
dc.identifier.urnURN:NBN:fi:aalto-2020123160455
dc.language.isoenen
dc.relation.ispartofACM SIGCHI annual conference on human factors in computing systemsen
dc.relation.ispartofseriesProceedings of the 2020 CHI Conference on Human Factors in Computing Systemsen
dc.rightsopenAccessen
dc.subject.keywordbiomechanical simulationen_US
dc.subject.keywordreinforcement learningen_US
dc.subject.keywordcomputational interactionen_US
dc.subject.keyworduser modelingen_US
dc.titlePredicting Mid-Air Interaction Movements and Fatigue Using Deep Reinforcement Learningen
dc.typeConference article in proceedingsfi
dc.type.versionpublishedVersion
Files