Compute-intensive perceptual reasoning tasks such as object detection and recognition are basic behaviorsthatservicerobots (Caoetal.,1995)mustperforminordertosupportandassisthumansover abroadvarietyofdifferentday-to-dayactivities. Effectiveobjectdetectionandrecognitionrequiresa greatdealofdatastorageandcomputepower,morethanwouldbetypicalofanembeddedrobotcontrolsystem. Ifitwerepossiblefortheembeddedsystemtooffloadthenecessarystorageandcompute capabilities to a more cost effective centralized infrastructure, it may be possible to introduce service robots with vastly improved perceptual reasoning capabilities at very low costs.
Thisideahasledustoconsidertheapplicabilityofcloudcomputing (Mell&Grance,2011;Armbrust etal.,2009)tosupportservicerobots,especiallyinperceptualreasoningtaskssuchasobjectdetection and recognition. There has been some research on solutions to open problems in the use of multiple networked robots, such as localization and mapping, cooperative robot learning, skills for service robots, and knowledge sharing. However, there is still a large gap between the abilities of individual or small groups of networked robots and what is required for truly useful multi-function autonomous service robots. Connecting multiple robots with a network helps us to share and pool resources, but architectures for Internet-scale distributed robot learning have not been extensively explored.
Thispaperpresentsacasestudyontheuseofacloudcomputingplatformtosupportroboticsapplications, allowing robots to offload heavy compute tasks such as machine vision to cloud infrastructure.
Compute-intensive perceptual reasoning tasks such as object detection and recognition are basic behaviorsthatservicerobots (Caoetal.,1995)mustperforminordertosupportandassisthumansover abroadvarietyofdifferentday-to-dayactivities. Effectiveobjectdetectionandrecognitionrequiresa greatdealofdatastorageandcomputepower,morethanwouldbetypicalofanembeddedrobotcontrolsystem. Ifitwerepossiblefortheembeddedsystemtooffloadthenecessarystorageandcompute capabilities to a more cost effective centralized infrastructure, it may be possible to introduce service robots with vastly improved perceptual reasoning capabilities at very low costs.
Thisideahasledustoconsidertheapplicabilityofcloudcomputing (Mell&Grance,2011;Armbrust etal.,2009)tosupportservicerobots,especiallyinperceptualreasoningtaskssuchasobjectdetection and recognition. There has been some research on solutions to open problems in the use of multiple networked robots, such as localization and mapping, cooperative robot learning, skills for service robots, and knowledge sharing. However, there is still a large gap between the abilities of individual or small groups of networked robots and what is required for truly useful multi-function autonomous service robots. Connecting multiple robots with a network helps us to share and pool resources, but architectures for Internet-scale distributed robot learning have not been extensively explored.
Thispaperpresentsacasestudyontheuseofacloudcomputingplatformtosupportroboticsapplications, allowing robots to offload heavy compute tasks such as machine vision to cloud infrastructure.
การแปล กรุณารอสักครู่..

Compute-intensive perceptual reasoning tasks such as object detection and recognition are basic behaviorsthatservicerobots (Caoetal.,1995)mustperforminordertosupportandassisthumansover abroadvarietyofdifferentday-to-dayactivities. Effectiveobjectdetectionandrecognitionrequiresa greatdealofdatastorageandcomputepower,morethanwouldbetypicalofanembeddedrobotcontrolsystem. Ifitwerepossiblefortheembeddedsystemtooffloadthenecessarystorageandcompute capabilities to a more cost effective centralized infrastructure, it may be possible to introduce service robots with vastly improved perceptual reasoning capabilities at very low costs.
Thisideahasledustoconsidertheapplicabilityofcloudcomputing (Mell&Grance,2011;Armbrust etal.,2009)tosupportservicerobots,especiallyinperceptualreasoningtaskssuchasobjectdetection and recognition. There has been some research on solutions to open problems in the use of multiple networked robots, such as localization and mapping, cooperative robot learning, skills for service robots, and knowledge sharing. However, there is still a large gap between the abilities of individual or small groups of networked robots and what is required for truly useful multi-function autonomous service robots. Connecting multiple robots with a network helps us to share and pool resources, but architectures for Internet-scale distributed robot learning have not been extensively explored.
Thispaperpresentsacasestudyontheuseofacloudcomputingplatformtosupportroboticsapplications, allowing robots to offload heavy compute tasks such as machine vision to cloud infrastructure.
การแปล กรุณารอสักครู่..
