A Novel Heavy Traffic Approach to Stochastic Optimal Control of Mobile Communications

Balancing high throughput with head-of-the-line wait time for queued data requires explicit stochastic process modeling of the time variations of the channel and the complex queuing process. However, such process models are too complicated for a direct solution. In this proposal, Intelligent Automation, Inc. (IAI) and its subcontractor, Prof. Robert Buche of the North Carolina State University, propose a novel heavy traffic approach to optimal stochastic control of mobile communications. Under the heavy traffic assumption and scaling, one can arrive at a limit system described by a stochastic differential equation (with reflection and/or jumps), which approximates the actual system. Based on the limit system, we can obtain the control by using the powerful novel numerical methods originally proposed by Kushner and Dupuis. Some other advantages of the proposed stochastic control design approach include: First, the controls obtained from the heavy traffic analysis are effective in the practical system, even operating far from the heavy traffic regime. Second, the heavy traffic modeling is general enough to handle a variety of configurations including TDMA, channel estimation, etc. Third, the proposed admission control design is independent of power control. Fourth, the proposed numerical methods sound and it has been theoretically justified by the convergence proofs.