Computer Science – Information Theory
Scientific paper
2007-04-07
Computer Science
Information Theory
Scientific paper
MIMO technology is one of the most significant advances in the past decade to increase channel capacity and has a great potential to improve network capacity for mesh networks. In a MIMO-based mesh network, the links outgoing from each node sharing the common communication spectrum can be modeled as a Gaussian vector broadcast channel. Recently, researchers showed that ``dirty paper coding'' (DPC) is the optimal transmission strategy for Gaussian vector broadcast channels. So far, there has been little study on how this fundamental result will impact the cross-layer design for MIMO-based mesh networks. To fill this gap, we consider the problem of jointly optimizing DPC power allocation in the link layer at each node and multihop/multipath routing in a MIMO-based mesh networks. It turns out that this optimization problem is a very challenging non-convex problem. To address this difficulty, we transform the original problem to an equivalent problem by exploiting the channel duality. For the transformed problem, we develop an efficient solution procedure that integrates Lagrangian dual decomposition method, conjugate gradient projection method based on matrix differential calculus, cutting-plane method, and subgradient method. In our numerical example, it is shown that we can achieve a network performance gain of 34.4% by using DPC.
Hou Thomas Y.
Liu Jia
No associations
LandOfFree
Cross-Layer Optimization of MIMO-Based Mesh Networks with Gaussian Vector Broadcast Channels does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Cross-Layer Optimization of MIMO-Based Mesh Networks with Gaussian Vector Broadcast Channels, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Cross-Layer Optimization of MIMO-Based Mesh Networks with Gaussian Vector Broadcast Channels will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-6904