Kenneth W. Shum

Sum Capacity of One-Sided Parallel Gaussian Interference Channels

The sum capacity of the one-sided parallel Gaussian interference channel is shown to be a concave function of user powers. Exploiting the inherent structure of the problem, we construct a numerical algorithm to compute it. Two suboptimal schemes are compared with the capacity-achieving scheme. One of the suboptimal schemes, namely iterative waterfilling, yields close-to-capacity performance when the cross link gain is small.

Fair Rate Allocation in Some Gaussian Multiaccess Channels

We can achieve all points in the capacity region of Gaussian multiple access channels by successive decoding and time-sharing. We discuss how to choose a particular point that is both Pareto optimal and fair to all users. The definition of our criterion of fairness is based on the theory of majorization. In economics, it is also known as the Lorenz order, which is used for measuring disparity in income distribution. We show that a unique solution according to such criterion exists in a large class of Gaussian multiple access channels. It turns out that the fair solution is the same as the well-known Nash bargaining solution. These two notions of fairness coincide due to the special structure of the capacity region. This provides a strong reason that we should pick it as the operational point. We also devise a fast algorithm that computes this point in some special cases

Rate Allocation for Cooperative Transmission in Parallel Channels

In this paper, we consider cooperative transmission between two source and destination pairs. Exchange of data is allowed between the two source nodes. In addition to the direct transmission link from the source to the intended destination, we have a two-hop relay link that sends the data via the neighboring source node. We partition the bandwidth into two parts, and each part is utilized by one source node, such that the transmissions from the two sources are orthogonal to each other. In this way, the cooperative interference channel is reduced to two independent broadcast channels. The bandwidth of each source node is divided into orthogonal sub-channels, and results from parallel broadcast channel is used to find the optimal allocation of power and rate to each links. We propose an iterative algorithm that maximizes the weighted sum rate, and plot the achievable rate region.

Transmitter cooperation by recycling dirty paper

The performance limit of transmitter cooperation in a wireless network with two source-destination pairs is investigated. We assume that each source node is equipped with a receiver and acts as a relay node to the other source node. A decode-and-forward half-duplex coding scheme based on a novel use of dirty-paper coding is devised. An outer bound is derived and compared with the achievable rate region. We show by numerical example that the gap between them can be quite small.

Iterative Waterfilling for Parallel Gaussian Interference Channels

We investigate synchronous iterative waterfilling power allocation algorithm for parallel Gaussian interference channels. Mobile terminals are allowed to update their powers in a fully distributed manner We show that a Nash equilibrium always exists in such a system. Some sufficient conditions for convergence are also derived.

Opportunistic Power Control with Rate Adaptation for Video Conferencing Services

We propose an opportunistic power control (OPC) algorithm with rate adaptation. It exploits channel variation and transmits opportunistically to optimize system performance. We show that it can be used to support delay-sensitive services like video conference applications. It works well when the wireless channel is changing moderately fast. For slowly varying channels, OPC yields good performance when dumb antenna is used. It outperforms the traditional target tracking approach in terms of both user capacity and power consumption.

To decode or to amplify: Mix and match for the two-way two-relay network

We study the problem of how two terminals exchange messages with the help of two relays. An upper bound and the achievable rate regions under amplify-forward (AF) and decode forward (DF) are derived. Then we mix these two strategies and propose three mixed strategies. We evaluate the performance of these strategies under different channel cases. It is shown that not only the basic strategies have their best performance in some scenarios, but also the mixed strategies match well under certain circumstances. Furthermore, some strategies can achieve or approach the upper bound under certain conditions.

The Capacity Region of Two-way Collision Networks

The collision channel without feedback is a system with multiple source nodes and a single destination node, where coordination among the source nodes and feedback from the destination node are not available. Its capacity region was first derived by Massey and Mathys. We generalize their model by introducing multiple destination nodes and relay nodes. In this paper, we consider linear collision networks in which the nodes lie on a straight line. The two nodes at the ends want to exchange data through the relay nodes in the middle. An outer bound on achievable rates is derived. By using a simple network code, we can show that all points within the outer bound are indeed achievable, and thus obtain the capacity region.