Joint admission and power control using branch & bound and gradual admissions

Power control has been extensively studied as an important way of mitigating interference and providing minimum signal to interference plus noise ratio (SINR) guarantees. Such formulation of power control is well-motivated in cellular PCS and UMTS, as both voice and streaming media require guaranteed short-term rates. A key difficulty is that the problem can easily become infeasible, implying that some link(s) must be dropped to accommodate the others. Since joint admission and power control is NP-hard, a host of heuristics have been proposed and implemented over the years, mostly based on the concept of gradual removals. More recently, the joint problem was revisited from a better-motivated Lagrangian relaxation / convex approximation viewpoint. In this contribution, we first derive a corresponding branch & bound algorithm that uses convex approximation for the bounding step. This can tackle moderate problem sizes, yielding optimal solution at much reduced average complexity relative to enumeration. Then, we propose a simple gradual admissions policy that appears promising. Simulations suggest that it can attain admission performance on a par with more complex methods, such as convex approximation, which are in turn known to outperform gradual removals.