On Multi-User Binary Computation Offloading in the Finite-Block-Length Regime

The Mobile Edge Computing framework provides small scale devices with the opportunity to offload their computational tasks to computing infrastructure at a network access point. Effective access to this infrastructure is contingent on the appropriate allocation of the available communication resources among the devices wishing to offload their tasks. In previous work, that allocation has been performed using guidance from classical characterizations of the fundamental limits on the rates at which reliable communication can be achieved, which are contingent on asymptotically long communication block lengths. However, the (latency) constraint on the time by which each device expects the results of its offloaded computational task imposes a natural limit on the block length. In this paper we show how a recent characterization of the rate limits in the finite-block-length regime can be incorporated into the problem of communication resource allocation for a K-device binary computational offloading system that employs the time-division multiple access (TDMA) scheme. We develop an efficient algorithm for that problem that is based on a tailored tree-search algorithm for the binary offloading decisions, a successive convex approximation algorithm for the transmission rates of the users, and closed-form solutions for the transmission powers and durations.