Research and Development of Stochastic Optimal Control Algorithms for Mobile Communications Systems

The overall objective of the effort (Phase I and Phase II) is to utilize Stochastic Optimal Control (SOC) techniques to derive and evaluate scheduling algorithms for Mobile Communication Systems. SOC techniques were recently employed to establish the optimality of the widely used Proportional Fair Sharing (PFS) scheduling algorithm under idealized conditions -- one single base station, unique service class and heavy traffic in all transmitters. This result provides an important link between well established SOC methods at one side and the practice of Mobile Communication Systems at the other side. We will use these developments as our starting point in Phase I, and investigate the possibility of extending these results when one or more of these assumptions are not met. Phase I will be focused in the scheduling problem only, using analytical techniques to: (1) investigate the performance of PFS under more general conditions, and (2) modifying PFS for improved performance. Limited numerical experiments will be performed to validate the theoretical developments. The most promising physical layer schemes derived in Phase I will be combined into a full protocol stack for High Rate Packet Data Systems in Phase II. We will focus our developments in the IS-856/CDMA2000 protocols, seeking the development of algorithms compatible with the emerging Third Generation Wireless Technologies. Northrop-Grumman Information Technology will be a subcontractor in the effort, providing expertise in the tactical operation of mobile networks. Professors Harold Kushner (Brown University), John Tsitsiklis (MIT) and Sanjoy Mitter (MIT) will serve as consultants, working with SSCI in the development of control theoretic methods for scheduler analysis and design.