Pareto Optimization of Data Transmission in a Partially Observed Communication Network

Abstract

The communication network under consideration consists of two stations: a single-server queuing system with non-stationary arrivals of packets and an unobserved stochastic channel with finite number of states. The service rate being proportional to the controlled transmission rate depends on the channel state. The probability of packet loss increases as the queue load becomes higher or the channel state worsens. The goal of the problem is to choose the transmission rate to minimize both packet losses and signal energy. The solution is based on visualization of Pareto optimal control policies with incomplete information. Two estimators (optimal filter and current queue state) and two operations (expectation and mode) are used to track changes in the stochastic channel. Computer simulation results are presented to compare the controls obtained.