Martingale Theory-Based Optimal Task Allocation in Heterogeneous Vehicular Networks

Abstract

Intelligent transportation systems (ITSs), which can fulfill people's increasing requirements on mobility, have arisen as a popular research area. In order to implement the ITSs, a large amount of data needs to be executed within requested time. Due to the remote locations, cloud centers are no longer sufficient in data off-loading. A heterogeneous vehicular network, which consists of three kinds of computation resources, i.e., a cloud center, an edge node, and multiple vehicles, is, therefore, constructed to solve the aforementioned issues. First, considering the heterogeneity of each computation resource, the delay bound can be derived based on the martingale theory. Especially, data off-loading to the cloud center or to the edge node is assisted by road side units (RSUs). The communication link from the source node to the cloud center or to the edge node is modeled as a two-hop one. By utilizing the min-plus algebra, the two-hop link can be abstracted and analyzed in an equivalent single system further. Then, given a requested delay time, the maximum off-loading capacity of each resource can be obtained from the derived delay bound. When the tasks are smaller than the network available off-loading capacity, an optimal task allocation problem can, therefore, be constructed to minimize the overall delay violation probability. The numerical results are presented to show the performance of the proposed task allocation scheme in the heterogeneous vehicular networks. In addition, it can be verified that the proposed optimal task allocation scheme performs better than the benchmarks in the literature.