Back-pressure-based adaptive routing algorithms where each packet is routed along a possibly different path have been extensively studied in the literature. However, such algorithms typically result in poor delay performance and involve high implementation complexity. In this paper, we develop a new adaptive routing algorithm built upon the widely studied back-pressure algorithm. We decouple the routing and scheduling components of the algorithm by designing a probabilistic routing table that is used to route packets to per-destination queues. The scheduling decisions in the case of wireless networks are made using counters called shadow queues. The results are also extended to the case of networks that employ simple forms of network coding. In that case, our algorithm provides a low-complexity solution to optimally exploit the routing–coding tradeoff.

The paper studies throughput-optimal scheduling in wireless networks when there are file arrivals and departures. In the case of single-hop traffic, the well-studied Max Weight algorithm provides soft priorities to links with larger queue lengths. If packets arrive in bursts during file arrival instants, then large variances in file sizes would imply that some links will have very large queue lengths while others will have small queue lengths. Thus, links with small queue lengths may be starved for long periods of time. An alternative is to use only MAC-layer queue lengths in making scheduling decisions; in fact, typically only this information is available since scheduling is performed at the MAC layer. Therefore the questions we ask in this paper are the following: (i) is scheduling using only MAC-layer queue length information throughput-optimal? and (ii) does it improve delay performance compared to the case where scheduling is performed using the total number of packets waiting at a link? We affirmatively answer both questions in the paper (the first theoretically and the second using simulations), making minimal assumptions on the transport-layer window control mechanism.

It has been conjectured that MWS-α scheduling policies with α going to zero are heavy-traffic optimal for scheduling in a generalized switch when the objective is to minimize the number of backlogged packets in the system. We examine this conjecture by first deriving optimal or heavy-traffic optimal policies for small switches and then comparing them to MWS- α policies by simulation. Our conclusion is that the conjecture is not true in general, i.e., there are simple topologies for which there exist policies that outperform MWS-α in heavy traffic.

We consider small generalized switches with less than or equal to four links, and study scheduling policies designed to minimize the total number of packets in the system. By focusing on very small switches, we are able to derive optimal or heavy-traffic optimal policies whose performance can then be compared to previously conjectured optimal policies. In particular, it has been conjectured that the max-weight policy with weight qalphais optimal in heavy-traffic when alpha rarr 0. Our results show that this conjecture is not true.

Multi-user MIMO downlink scenario has been paid considerable attentions in recent years for its potential large capacity. However, when there are not enough antennas at the transmitter, making all the users available may not be a good solution. The situation can be improved by selecting the right users among all users at the receiver and meanwhile a maximum capacity can be achieved. In this paper, we use the correlation of channel matrices of users and develop an algorithm to select the right users with low complexity to achieve the total capacity of the whole system. Moreover, we design a strategy to adaptively select receive antennas of users at the receiver which can further increase the capacity of the MIMO system.