A Generalized Nash Game for Mobile Edge Computation Offloading

The number of tasks performed on wireless mobile devices instead of stationary computers is steadily increasing. Due to the limited resources of battery and computation capability on these devices, computation offloading became a relevant concept. We consider several mobile users with a splittable computation task each, which try to minimize their own computation time. All of them are connected to a central access point, where a cloudlet with limited computation power can be utilized for offloading fractions of their tasks. To account for the selfishness of the users and their individual goals, we propose a game theoretic framework resulting in a nonconvex generalized Nash game. The decision of each mobile user, which fraction of his task to offload, depends on the offloading decisions of the others, since they share the communication resources and computation capabilities of the cloudlet. We prove the existence and uniqueness of the generalized Nash equilibrium and propose an algorithm for computing it efficiently. In addition, we show that the price of anarchy of our model is one, which is the best possible value, and investigate the advantage of computation offloading numerically. Furthermore, we extend our model to a scenario, where mobile users are able to offload parts of their computation in repeated sessions during a given time period. This allows every mobile user to offload multiple successive tasks.