TY - GEN
T1 - Edge-to-edge cooperative artificial intelligence in smart cities with on-demand learning offloading
AU - Zhang, Li
AU - Wu, Jun
AU - Mumtaz, Shahid
AU - Li, Jianhua
AU - Gacanin, Haris
AU - Rodrigues, Joel J.P.C.
N1 - Funding Information:
ACKNOWLEDGMENT This work is supported by National Natural Science Foundation of China (Grant No. 61431008 and 61571300), by National Funding from the FCT - Fundac¸ão para a Ciência e a Tecnologia through the UID/EEA/50008/2019 Project; by RNP, with resources from MCTIC, Grant No. 01250.075413/2018-04, under the Centro de Referência em Radiocomunicac¸ões - CRR project of the Instituto Nacional de Telecomunicac¸ões (Inatel), Brazil and by Brazilian National Council for Research and Development (CNPq) via Grant No. 309335/2017-5.
Publisher Copyright:
© 2019 IEEE.
PY - 2019/12
Y1 - 2019/12
N2 - With the development of smart cities, the demand for artificial intelligence (AI) based services grows exponentially. The existing works just focus on cloud- edge or edge-device cooperative AI which suffers low learning efficiency of AI, while edge-to-edge cooperative AI is still an unresolved issue. Moreover, the existing researches concentrate on the computation offloading of the AI-based task, ignoring that it is a brain-like task performing sophisticated processing to raw data, which leads to the high latency and low quality of the learning services. To address these challenges, this paper proposes an on-demand learning offloading mechanism for edge-to-edge cooperative AI. Firstly, the principle of the learning capability and its offloading are proposed for the formal description of the learning resources migration. Secondly, the proposed mechanism realizes the bilateral learning offloading utilizing edge-to-edge and cloud-edge collaborations to handle AI-based tasks with high learning efficiency and resource utilization rate. Moreover, we model the edge-to-edge learning offloading allocation based on the concatenation of deep neural network (DNN) subtasks and their heterogeneous requirement of learning resources. Simulation results indicate the rationality and efficiency of the proposed mechanism.
AB - With the development of smart cities, the demand for artificial intelligence (AI) based services grows exponentially. The existing works just focus on cloud- edge or edge-device cooperative AI which suffers low learning efficiency of AI, while edge-to-edge cooperative AI is still an unresolved issue. Moreover, the existing researches concentrate on the computation offloading of the AI-based task, ignoring that it is a brain-like task performing sophisticated processing to raw data, which leads to the high latency and low quality of the learning services. To address these challenges, this paper proposes an on-demand learning offloading mechanism for edge-to-edge cooperative AI. Firstly, the principle of the learning capability and its offloading are proposed for the formal description of the learning resources migration. Secondly, the proposed mechanism realizes the bilateral learning offloading utilizing edge-to-edge and cloud-edge collaborations to handle AI-based tasks with high learning efficiency and resource utilization rate. Moreover, we model the edge-to-edge learning offloading allocation based on the concatenation of deep neural network (DNN) subtasks and their heterogeneous requirement of learning resources. Simulation results indicate the rationality and efficiency of the proposed mechanism.
KW - Edge AI
KW - Edge-to-edge collaboration
KW - Learning offloading
UR - http://www.scopus.com/inward/record.url?scp=85081951960&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85081951960&partnerID=8YFLogxK
U2 - 10.1109/GLOBECOM38437.2019.9013878
DO - 10.1109/GLOBECOM38437.2019.9013878
M3 - Conference contribution
AN - SCOPUS:85081951960
T3 - 2019 IEEE Global Communications Conference, GLOBECOM 2019 - Proceedings
BT - 2019 IEEE Global Communications Conference, GLOBECOM 2019 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2019 IEEE Global Communications Conference, GLOBECOM 2019
Y2 - 9 December 2019 through 13 December 2019
ER -
