Joint Computation Offloading and Radio Resource Allocations in Wireless Cellular Networks

Abstract

Mobile computation offloading (MCO) is an emerging technology to migrate resource-intensive computations from resource-limited mobile devices to resource-rich devices (such as a cloud server) via wireless access. For applications that are time sensitive, offloading to nearby cloudlets is preferred, compared to offloading to a remote cloud server, in order to save the data transmission delay. On the other hand, the limited computing capabilities and the wireless transmission conditions to access the cloudlet servers can both affect the offloading performance, especially when multiple users are competing for offloading services. In this paper, we study joint computation offloading and radio resource allocations in small cell cellular systems, where cloudlet servers are colocated at the base stations. Our objective is to minimize the total energy consumption of the system, for both data transmissions and task executions, subject to the hard latency requirements of the applications. The problem is first formulated as a mixed integer nonlinear optimization problem, and then decomposed into multiple power allocation subproblems and an offloading decision subproblem. The power allocation subproblems are non-convex, which are reformulated and solved iteratively. Their results are fed into the offloading decision subproblem, which then becomes a linear integer (bi- nary) problem, and can be converted into a matching problem and solved using a modified Kuhn-Munkres (K-M) algorithm. Simulation results demonstrate that the joint optimization can significantly improve the offloading efficiency, compared to other resource allocation methods.