Distributed Backlog-Driven Power Control in Wireless Networking

We address the problem of distributed power control for supporting packetized traffic in wireless ad hoc networks (e.g. 802.11 based wireless LANs). The interference experienced by a link coexisting with other links in a shared wireless medium is responsive to the actions of the transmitter on the link. Consequently, the evolution of seemingly independent queues at autonomously acting transmitters is tightly entangled through the shared wireless channel. Thus, an exact analysis of the distributed power control problem which also incorporates queuing dynamics is intractable. To establish a performance benchmark, we first design the optimal centralized backlog aware power control algorithm (Oracle) in a dynamic programming (DP) framework. We then propose a heuristic backlog aware distributed power control algorithm (BDD). The implementation of BDD is based on randomized selection from lookup tables which are computed offline at each transmitter. The computational complexity and memory requirements for BDD are independent of the network size and topology, making it attractive from a practical perspective. Experimental results demonstrate that BDD closely matches Oracle in performance and enhances system throughput by 20-30% compared to benchmark backlog insensitive power control algorithms (e.g. Foschini-Miljanic).