Adaptive CSMA under the SINR Model: Efficient Approximation Algorithms for Throughput and Utility Maximization

TL;DR

This paper introduces efficient local approximation algorithms for adaptive CSMA in wireless networks under the SINR model, enabling fast and accurate estimation of attempt rates (fugacities) for throughput and utility maximization.

Contribution

It presents novel local algorithms that support desired service rates and optimize fugacities with convergence rates independent of network size, outperforming existing stochastic methods.

Findings

Algorithms achieve fast convergence to near-optimal fugacities.

Proposed methods outperform stochastic gradient descent in speed.

Algorithms support utility maximization with high accuracy.

Abstract

We consider a Carrier Sense Multiple Access (CSMA) based scheduling algorithm for a single-hop wireless network under a realistic Signal-to-interference-plus-noise ratio (SINR) model for the interference. We propose two local optimization based approximation algorithms to efficiently estimate certain attempt rate parameters of CSMA called fugacities. It is known that adaptive CSMA can achieve throughput optimality by sampling feasible schedules from a Gibbs distribution, with appropriate fugacities. Unfortunately, obtaining these optimal fugacities is an NP-hard problem. Further, the existing adaptive CSMA algorithms use a stochastic gradient descent based method, which usually entails an impractically slow (exponential in the size of the network) convergence to the optimal fugacities. To address this issue, we first propose an algorithm to estimate the fugacities, that can support a given set of desired service rates. The convergence rate and the complexity of this algorithm are independent of the network size, and depend only on the neighborhood size of a link. Further, we show that the proposed algorithm corresponds exactly to performing the well-known Bethe approximation to the underlying Gibbs distribution. Then, we propose another local algorithm to estimate the optimal fugacities under a utility maximization framework, and characterize its accuracy. Numerical results indicate that the proposed methods have a good degree of accuracy, and achieve extremely fast convergence to near-optimal fugacities, and often outperform the convergence rate of the stochastic gradient descent by a few orders of magnitude.