Task Offloading in Mobile Edge Computing: Intractability and Proposed Approaches

The problem of task offloading in mobile edge computing is recently known to be NP-hard. We identify in this paper several practical assumptions, for reducing the problem into its restricted variants. We however find that task offloading remains to be intractable (NP-complete) even in the context where the number of mobile users for a given channel is fixed, and vice versa. Accordingly in this paper we position several approaches (e.g., SAT/SMT encodings, task offloading as stable matching, deep reinforcement learning, and general machine learning approaches) as means to tackle the problem.