Available Bit Rate Teletraffic Control Using Model Predictive Control

Model Predictive Control is analyzed as a strategy for coping with the unknown, variable and non-deterministic action and measurement delays which can occur in congestion control of available-bit-rate teletraffic control for highspeed data networks. It is shown that the state-feedbackplus-estimator certainty equivalence nature of this control law is particularly amenable to coping with these delays. Both the estimator update rule and the receding-horizon controller can be modified to accommodate absent network measurements. This is standard for the estimator. But involves a novel approach to the receding-horizon control law, which for the congested network results in diminishing the network utilization objective in favor of queue stability.