Distributed Asynchronous Optimization Framework for the MISO Interference Channel

We study the distributed optimization of transmit strategies in a multiple-input, single-output (MISO) interference channel (IFC). Existing distributed algorithms rely on strictly synchronized update steps by the individual users. They require a global synchronization mechanism and potentially suffer from the synchronization penalty caused by e.g., backhaul communication delays and fixed update sequences. We establish a general optimization framework that allows asynchronous update steps. The users perform their computations at arbitrary instants of time and do not wait for information that has been sent to them. Based on certain bounds on the amount of asynchronism that is present in the execution of the algorithm, we are able to characterize its convergence. As illustrated by our numerical results, the proposed algorithm is not excessively slowed down by neither communication delays, nor by specific update orders, and thus enables faster convergence to (local) optimal solution.