Cost-Effective V2X Task Offloading in MEC-assisted Intelligent Transportation Systems
Intelligent Transportation Systems (ITS) will become an essential part of every city in the near future. They should support various vehicle-to-everything (V2X) applications that improve road safety or even enable autonomous driving. Recently, the European Telecommunications Standards Institute (ETSI) introduced a multi-access (mobile) edge computing concept as a promising solution to satisfy the V2X delay and computational requirements. Based on this concept, the tasks generated by V2X applications can be offloaded to servers at the edge of the radio access network (RAN). There is a need for a task offloading algorithm that minimizes the ITS operator expenses connected with the servers deployment and maintenance while satisfying the requirements of the V2X applications. Most of the existing papers in the literature do not pay much attention to queuing delays at servers. In this paper, the queuing delays are analyzed by considering a general-type task computational time distribution. A non-linear optimization problem is formulated to minimize the ITS operator expenses subject to delays and computational resources constraints. The flexibility is also improved by considering that a delay constraint is satisfied with a given probability. To solve this problem, a method for linearization of the problem is proposed, and consequently, an algorithm based on Integer Linear Programming (ILP) is designed. A heuristic algorithm called Costeffective Heuristic Algorithm for Task offloading (CHAT) is also introduced that provides close to optimal results and has much lower computational complexity than the ILP algorithm. The efficiency of the CHAT algorithm is studied in several scenarios in terms of the computational time, delays, and the total server energy consumption as the cost function. The results show that the CHAT algorithm satisfies the requirements of the V2X applications in all the considered scenarios and reduces the ITS operator expenses over twice compared with other algorithms proposed in the literature.