Lightweight power control for energy-utility optimization in wireless networks

We consider an interference-limited, ad-hoc wireless network in the high SINR regime and address the optimization of network utility and energy efficiency by cross-layer network control. Unlike the typical complex approach that requires solving a scheduling, routing and power control problem at each time slot, we propose running a single iteration of a gradient power control algorithm towards the optimal power allocation, together with backpressure multipath routing and flow control. Despite the fact that the respective optimizations at each time slot are never fully solved, we prove, under a high SINR assumption, that the proposed updates suffice to optimize network utility and energy efficiency. Main components of the joint algorithm are flow control at each node (based on local queues), backpressure routing/scheduling, and power control driven by backlog, interference, and power cost related information. We provide simulation results that illustrate the convergence to the optimal flow rates and link powers, compare against related algorithms from the literature, and examine the validity of the high SINR approximation. Our approach may allow in-practice performance gains and inspire more research on low-complexity, practical network control.