Date: Wednesday, November 1, 2006
Special University Ph.D. Oral Examination
Time: 3:00 pm
Location: Packard #202
< back to top
Multi-User Bit-Loading: Theoretical Limits and Practical Implementations
David Daxiao Yu
of Electrical Engineering, Stanford University
Resource allocation is the key to the performance of multi-user networks, where users compete for resources because of interference, and inevitably cooperate to share resources in order to optimize certain utility functions such as weighted sum rate of all users. Resource allocation is also subject to practical constraints such as finite channel partitioning, discrete constellations and code rates.
In this talk we will discuss optimal resource allocation in Gaussian Multiple Access Channel (MAC) and discrete resource allocation in Gaussian interference channels. Characterization of capacity region/achievable region will be described, and efficient multi-user bit-loading algorithms for both continuous and discrete cases will be introduced. For continuous bit-loading, special structures of the KKT conditions are exploited, and the efficiency of the block coordinate ascending method is readily shown in solving both Single Input Single Output (SISO) MAC and Single Input Multiple Output (SIMO) MAC weighted sum rate maximization problems. For discrete bit-loading in Gaussian interference channels, the non-decreasing property of rounding an interference function is the key to the design of efficient algorithms that exhibit linear complexity in the number of users.