Efficient task processing in heterogeneous vehicular networks is one of the key enablers of future intelligent transportation systems (ITSs). In this paper, first, we present a data uploading and task offloading scenario in heterogeneous vehicular networks, by which vehicles can offload tasks to peer vehicles, edge nodes and cloud nodes to reduce task processing delay and overall energy consumption. Then, we formulate a Cooperative Data Uploading and Task Offloading (CDUTO) problem, which aims at minimizing average task service delay and system energy consumption simultaneously by jointly considering common data requirement of tasks, heterogeneous resources and vehicle mobility. On this basis, we transform CDUTO into a Markov chain and propose a problem-specific Markov-Approximation algorithm (MAA) to solve CDUTO. In particular, specific initialization strategy is designed to initialize the state of Markov chain, and the uploading strategy is designed to decide the uploading vehicle of common data. The offloading decision is made when the Markov chain reaches stationary distribution. We prove that the gap of MAA is $\frac{1}{\rho}\log\vert \Phi\vert$, where $\rho$ is a constant and $\Phi$ is the size of solution space. Finally, we build the simulation model and give comprehensive performance evaluation, which demonstrates that MAA outperforms the baseline algorithm for around 29%.
Cooperative Data Uploading and Task Offloading via Markov-Approximation in Heterogeneous Vehicular Networks
2023-09-24
806542 byte
Conference paper
Electronic Resource
English