Task offloading is critical to enable various computation intensive applications in vehicular networks as well as accelerate the intellectualization of transportation system. In this paper, we first present a task offloading architecture in heterogeneous vehicular networks via end-edge-cloud cooperation. In particular, we investigate the scenario where each task may consist of multiple dependent subtasks, which have to be processed in sequence and in time. Then, a Dependency-Aware Task Offloading Problem (DATOP) is formulated as a Mixed Integer Nonlinear Programming problem (MINLP), aiming at minimizing the average service delay of tasks. Further, we design an offline method as a benchmark, which linearizes the problem into a Mixed Integer Linear Programming problem (MILP) and obtains the optimal solution through Lingo. Moreover, we propose a three-stage heuristic algorithm, which computes the critical path based on the derived task graph and schedules subtasks on the critical path via dynamic programming, and then the subtasks on non-critical paths are scheduled via a greedy strategy. Finally, we build the simulation model and conduct comprehensive performance evaluation, which demonstrates the effectiveness of the proposed online algorithm.
Dependency-aware Task Offloading via End-Edge-Cloud Cooperation in Heterogeneous Vehicular Networks
2022-10-08
641053 byte
Conference paper
Electronic Resource
English