Lyapunov-based Dynamic Computation Offloading Optimization in Heterogeneous Vehicular Networks

Abstract

As a flying Mobile Edge Computing (MEC) server, the Unmanned aerial vehicle (UAV) has been employed to strength the computation capability of vehicular networks. However, the intermittent connection between moving vehicles and UAVs, and unknown distribution of computation requests bring great challenges to the online computation offloading optimization. In this work, we propose a dynamic vehicular computation offloading problem with hybrid Vehicle-to-Vehicle (V2V), Vehicle-to-Roadside unit (V2R) and Vehicle-to-UAV (V2U) communications. In order to minimize the long-term computation offloading delay in dynamic environment, we present a Lyapunov-based dynamic computation offloading (LDCO) algorithm, which transforms the original problem into a series of subproblems by minimizing the derived upper bound of the Lyapunov drift-plus-penalty function. Each subproblem is then formulated as a two-dimensional multiple knapsack problem (TDMKP), which only involves the information of current vehicles' positions and computation requests at each time slot. Comprehensive studies show significant performances of the proposed computation offloading architecture together with the dynamic offloading algorithms.