COIN: Generalized MaxWeight Algorithms over Non-Convex Rate Regions in Wireless Networks with Multiple Antennas


The holy grail of network information theory is to compute the capacity region of wireless networks: however, it is questionable whether this goal can be achieved soon. Therefore, information theorists and communication engineers increasingly favor simpler but more realistic goals, such as finding approximate expressions for the capacity region within predictable error margins. An intriguing approach in this direction is the linear deterministic model (LDM) recently developed by Avestimehr et al [ADT11]. LDM embraces characteristic wireless channel features such as signal strength, broadcast, multiple access and introduces deterministic bit pipes between network entities. In this research proposal a joint approach to wireless network design using LDM together with dynamic network control shall be explored. The main goals are: 1. to make LDM available for cellular wireless network design. Recent results on LDM cellular-type channels shall be adopted and extended toward more realistic schemes which include multiple cells, transceiver cooperation, relaying, etc. 2. to bring LDM capacity-achieving schemes into a LDM network utility maximization framework and to devise efficient combinatorial algorithms. 3. to adopt and extend existing control strategies for dynamic cellular LDM networks and transfer the results to Gaussian networks. One of the main envisioned results of this project is to clearly indicate whether advanced interference-aware strategies in cellular networks, i.e. those which do not treat interference as noise but detect and subtract interference, will pay off in terms of performance gains accommodating for investment/operational costs.