Adaptive Data Sharing and Computation Offloading in Cloud-Edge Computing with Resource Constraints.

Wenjie Chu,Haiyan Zhao,Zhi Jin,Zhenjiang Hu

DOI: https://doi.org/10.1109/SMC42975.2020.9283207

2020-01-01

Abstract:Collaborative tasks require the participation of multiple agents. Each agent in collaboration needs sufficient data to make optimal decisions. However, in general, each agent can only collect and process a limited amount of data due to resource constraints. Peer-to-peer data sharing can enrich local observations, but a particular agent may not have enough resources to adequately store and process data, thus compromising group decision making. Cloud-Edge Computing (CEC) can relieve agents of these limitations by providing them with further storage and computing resources through connected cloud-like infrastructures. However, CEC-based collaborations currently face two key challenges: 1) lack of adaptability to resource restrictions in data sharing; 2) no support of offloading non-trivial tasks with complex data dependencies. This paper proposes an approach to realize adaptive data sharing and support computation offloading. Roughly speaking, the paired parameterized-structure is designed based on data flow analysis and bidirectional transformations to benefit adaptive data synchronization and offloading. And a hybrid offloading mechanism is offered for allocating computations among agents and the cloud, regarding data dependencies and restrictions. We demonstrate the feasibility and flexibility through a collaborative victim search and rescue case. Experiments show that our approach outperforms state-of-the-art methods.